dsfsi/za-mafoko-ai-glossary · Datasets at Hugging Face

id string	eng string	afr string	xho string	zul string	eng_pos_or_descriptor string	afr_pos_or_descriptor string	xho_pos_or_descriptor string	zul_pos_or_descriptor string
google_research_ai_terminology_0	Accuracy	akkuraatheid	Ukuchaneka	Ukunemba	It is the proportion of all classifications that were correct, whether positive or negative. Emails example: In the spam classification example, accuracy measures the fraction of all emails correctly classified.	Dit is die gedeelte van alle klassifikasies wat korrek was, ongeag of dit positief of negatief was. E-posvoorbeeld: In die strooiposklassifikasievoorbeeld meet akkuraatheid die breukdeel van alle e-posse wat korrek geklassifiseer is.	Ngumlinganiselo wazo zonke iindidi zokuhlela ebezichanekile, nokuba bezilungile okanye zingalunganga. Umzekelo we-imeyile: Kumzekelo wokuhlelwa kwe-spam, ukuchaneka kulinganisa iqhezu lazo zonke ii-imeyile ezihlelwe ngokuchanekileyo.	Kuyisilinganiso sakho konke ukuhlunga obekunembile, kungakhathaliseki ukuthi kuhle noma kubi. Isibonelo sama-imeyili: Esibonelweni sokuhlungwa kukagaxekile, ukunemba kulinganisa ingxenye yawo wonke ama-imeyili ahlungwe ngokunembile.
google_research_ai_terminology_1	Active learning	aktiewe leer	Ukufunda ngokusebenza	Ukufunda okusebenzayo	A special case of machine learning in which a learning algorithm can interactively query a human user (or some other information source), to label new data points with the desired outputs.	’n Spesiale geval van masjienleer waarin ’n leeralgoritme interaktief navraag by ’n menslike gebruiker (of ’n ander inligtingbron) kan doen om nuwe datapunte met die verlangde uitsette te etiketteer.	Imeko ekhethekileyo yokufunda koomatshini apho i-algorithm yokufunda inokubuza ngokudibeneyo kumsebenzisi ongumntu (okanye omnye umthombo wolwazi), ukulebheyila amanqaku amatsha edatha kunye neziphumo ezifunwayo.	Isimo esikhethekile sokufunda ngomshini lapho i-algorithm yokufunda ingabuza ngokuxhumana nomsebenzisi ongumuntu (noma omunye umthombo wolwazi), ukulebula amaphoyinti edatha amasha ngemiphumela efiselekayo.
google_research_ai_terminology_2	AI alignment	KI-belyning	Ulungelelwaniso lweAI	Ukuqondanisa kwe-AI	Aims to steer AI systems toward a person's or group's intended goals, preferences, and ethical principles. An AI system is considered aligned if it advances the intended objectives. A misaligned AI system pursues unintended objectives.	Dit is daarop gemik om AI-stelsels in die rigting van ’n persoon of groep se bedoelde doelwitte, voorkeure en etiese beginsels te stuur. ’n AI-stelsel word as gerig beskou as dit die bedoelde doelwitte bevorder. ’n Wangerigte AI-stelsel streef onbedoelde doelwitte na.	Ijolise ekuqhubeni iinkqubo ze-AI ukuya kwiinjongo ezijoliswe kumntu okanye iqela, ukhetho, nemigaqo yokuziphatha. Inkqubo ye-AI ithathwa njengelungelelanisiweyo ukuba iqhubela phambili iinjongo ekujoliswe kuzo. Inkqubo ye-AI engalunganga ilandela iinjongo ezingalindelekanga.	Imigomo yokuqondisa amasistimu we-AI emigomweni ehloselwe umuntu noma iqembu, okuncanyelwayo, nemithetho yokuziphatha. Isistimu ye-AI ibhekwa njengeqondanisiwe uma ithuthukisa imigomo ehlosiwe. Isistimu ye-AI engaqondanisiwe kahle iya emigomweni engahlosiwe.
google_research_ai_terminology_3	Algorithmic bias	algoritmesydigheid	Ukuthatha icala kwealgorithm	Ukuchema ngokwe-algorithmic	A systematic and repeatable errors in a computer system that create "unfair" outcomes, such as "privileging" one category over another in ways different from the intended function of the algorithm.	’n Sistematiese en herhaalbare fout in ’n rekenaarstelsel wat “onregverdige” uitkomste skep, soos om een kategorie bo ’n ander te “bevoordeel” op maniere wat van die bedoelde funksie van die algoritme afwyk.	Iimpazamo ezicwangcisiweyo neziphindaphindwayo kwinkqubo yekhompyutha edala iziphumo "ezingalunganga", ezifana "nokubeka phambili" olunye udidi ngaphezulu kolunye ngeendlela ezahlukileyo kumsebenzi ocetyiweyo we-algorithm.	Amaphutha wesistimu naphindekayo kusistimu yekhompyutha asungula imiphumela "engafanele", "njengokuphakamisa" isigaba esithile ngaphezu kwesinye ngezindlela ezihlukile emsebenzini obuhlosiwe we-algorithm.
google_research_ai_terminology_4	Algorithmic transparency	algoritmedeursigtigheid	Ukubonakala kwealgorithm	Ukungafihlwa ngokwe-algorithmic	The principle that the factors that influence the decisions made by algorithms should be visible, or transparent, to the people who use, regulate, and are affected by systems that employ those algorithms.	Die beginsel dat die faktore wat die besluite beïnvloed wat deur algoritmes geneem word, sigbaar, of deursigtig, moet wees vir die mense wat stelsels wat daardie algoritmes benut, gebruik en reguleer en daardeur geraak word.	Umgaqo wokuba izinto ezichaphazela izigqibo ezenziwe ziialgorithm kufuneka zibonakale, okanye zibe lubala kubantu abazisebenzisayo, abazilawulayo, nabachaphazelekayo ziinkqubo ezisebenzisa ezo zixhobo.	Umthetho wokuthi izici ezinegalelo ezinqumweni ezenziwa ama-algorithms kufanele zibonakale, noma zingafihleki, kubantu abasebenzisa, baqondise, futhi bathintwe amasistimu asebenzisa lawo ma-algorithm.
google_research_ai_terminology_5	Artificial Intelligence	kunsmatige intelligensie	Ubungqondi bekhompyutha	Ubuhlakani Bokwakhiwe	AI / In its broadest sense, is intelligence exhibited by machines, particularly computer systems. It is a field of research in computer science that develops and studies methods and software that enable machines to perceive their environment and use learning and intelligence to take actions that maximize their chances of achieving defined goals.	KI / Dit is in die wydste betekenis intelligensie wat deur masjiene, veral rekenaarstelsels, getoon word. Dit is ’n navorsingsveld in rekenaarwetenskap wat studiemetodes en sagteware ontwikkel wat dit vir masjiene moontlik maak om hul omgewing waar te neem en leer en intelligensie te gebruik om handelinge uit te voer wat hul kans so groot moontlik maak om gedefinieerde doelwitte te bereik.	AI / Xa ichazwa ngeyona ndlele igabalala, bubukrelekrele obubonakaliswa ngoomatshini, ingakumbi iinkqubo zekhompyutha. Licandelo lophando kwisayensi yekhompyutha ephuhlisa ize ifunda iindlela nesoftware eyenza ukuba oomatshini babone okuwungqongileyo kwaye usebenzise ukufunda nobukrelekrele ukuthatha iintshukumo ezandisa amathuba awo okufikelela kwiinjongo ezichaziweyo.	Ngomqondo obanzi, ukuhlakanipha okubonakaliswa imishini, ikakhulu amasistimu ekhompyutha. Iyinkundla yocwaningo lwesayensi yekhompyutha esungula futhi icwaninge izindlela nesofthiwe evumela imishini ukuba ibone indawo futhi isebenzise ukufunda nokuhlakanipha ukuze ithathe izinyathelo ezikhuphula amathuba ayo okufinyelela imigomo ebekiwe.
google_research_ai_terminology_6	Artificial Neural Network	kunsmatige neurale netwerk	INethiwekhi yeNgqondo Eyenziweyo	Ubuxhakaxhaka Bokunikezelana Obakhiwe	ANN / In machine learning, a neural network (also artificial neural network or neural net, abbreviated ANN or NN) is a model inspired by the structure and function of biological neural networks in animal brains. An ANN consists of connected units or nodes called artificial neurons, which loosely model the neurons in the brain.	KNN / In masjienleer is ’n neurale netwerk (ook kunsmatige neurale netwerk of neurale net, afgekort as KNN of NN) ’n model wat geïnspireer is deur die struktuur en funksie van biologiese neurale netwerke in die breine van diere. ’n KNN bestaan uit gekoppelde eenhede of nodusse genaamd kunsmatige neurone, wat losweg na die neurone in die brein gemodelleer is.	ANN / Ekufundweni koomatshini, i-neural nethwekhi (ne-artificial neural nethwekhi okanye i-neural net, i-ANN xa ifinyeziwe okanye i-NN) yimodeli engumzekelo ekhuthazwa sisakhiwo nomsebenzi we-biological neural networks kubuchopho bezilwanyana. I-ANN iqulathe iiyunithi ezidityanisiweyo okanye iindawo ezibizwa ngokuba zii-neuron ezenziweyo, ezibonisa ngokukhululekileyo imithambo-luvo ebuchotsheni.	Ekufundeni komshini, ineural network (ebizwa nangokuthi yi-artificial neural network noma ineural net, enesifinyezo esithi ANN noma NN) iyimodeli esekelwa ukwakheka nokusebenza kwamabiological neural network ebuchosheni bezilwane. I-ANN iqukethe amayunithi noma amanodi axhumene abizwa ngokuthi ama-artificial neuron, alingisa amaneuron asebuchosheni.
google_research_ai_terminology_7	Autonomous Vehicles	outonome voertuig	Izithuthi ezizilawulayo	Izimoto Ezizishayelayo	Refers to a car that is capable of operating with reduced or no human input. Synonyms: self-driving car, driverless car, robotaxi, robotic car or robo-car.	Dit verwys na ’n motor wat in staat is om met verminderde of geen menslike insette nie te werk. Sinonieme: selfbestuurmotor, bestuurderlose motor, robottaxi, robotiese motor of robotmotor.	Ibhekisa kwimoto ekwaziyo ukusebenza ngegalelo lomntu elinciphileyo okanye kungekho galelo lomntu. Izithethantonye: imoto eziqhubayo, imoto engenamqhubi, i-robotaxi, imoto yerobhothi okanye i-robo-car.	Zibhekisela emotweni ekwazi ukuhamba ngokuqondisa okuncane komuntu noma ngaphandle kwako. Amagama afanayo: imoto ezishayelayo, imoto engenamshayeli, irobotaxi, imoto eyirobhothi noma irobo-car.
google_research_ai_terminology_8	Bias-variance tradeoff	sydigheid-variansie-afruiling	Ukuthatha icala kwiyantlukwano yorhwebo	Ilumbanisamaphutha	Describes the relationship between a model's complexity, the accuracy of its predictions, and how well it can make predictions on previously unseen data that were not used to train the model.	Dit beskryf die verhouding tussen ’n model se kompleksiteit, die akkuraatheid van die voorspellings daarvan, en hoe goed dit voorspellings kan doen op grond van data wat nie vroeër gesien is nie en nie gebruik is om die model op te lei nie.	Ichaza unxulumano phakathi kokuntsokotha kwemodeli, ukuchaneka koqikelelo lwayo, nendlela enokwenza ngayo uqikelelo kwidatha ebingabonwa ngaphambili engazange isetyenziswe ekuqeqesheni imodeli.	Ichaza ubuhlobo phakathi kobunkimbinkimbi bemodeli, ukunemba kokuqagela kwayo, nendlela ekwazi ukuqagela kahle ngayo kudatha engakaze ibonwe ngaphambilini engazange isetshenziswe ekuqeqesheni leyo modeli.
google_research_ai_terminology_9	Chatbot	kletsbot	Ichatbot	Ifuzelankulumo	A software application or web interface that is designed to mimic human conversation through text or voice interactions. Also spelled as chat bot.	’n Sagtewareapp of webkoppelvlak wat ontwerp is om menslike gesprek deur teks- of steminteraksies na te boots. In Engels gespel as “chatbot” of “chat bot”.	Iaplikheyshini yesoftware okanye ujongano lwewebhu eyilelwe ukulinganisa incoko yabantu ngokubhaliweyo okanye ukusetyenziswa kwelizwi. Ikwapelwe ngokuthi chat bot.	I-application yesofthiwe noma isixhumi esibonakalayo sewebhu esiklanyelwe ukulingisa ingxoxo yabantu ngombhalo noma ngokuxhumana ngezwi. Iphinde ibhalwe nangokuthi chat bot.
google_research_ai_terminology_10	Computer vision	rekenaarvisie	Umbono wekhompyutha	Umbono wekhompyutha	An interdisciplinary field that deals with how computers can be made to gain high-level understanding from digital images or videos. From the perspective of engineering, it seeks to automate tasks that the human visual system can do. As a scientific discipline, computer vision is concerned with the theory behind artificial systems that extract information from images.	’n Interdissiplinêre vakgebied wat gaan oor hoe rekenaars gemaak kan word om hoëvlakbegrip van digitale prente of video’s te kry. Uit ’n ingenieursoogpunt is dit daarop gemik om take te outomatiseer wat die menslike visuele stelsel kan verrig. As ’n wetenskaplike dissipline is rekenaarvisie gemoeid met die teorie agter kunsmatige stelsels wat inligting uit prente haal.	Inkalo yezifundo ezahlukeneyo ejongene nendlela iikhompyutha ezinokuthi zenziwe ngayo ukufumana ukuqonda okuphezulu kwimifanekiso yedijithali okanye iividiyo. Ngokwembono yobunjineli, ifuna ukuzenzela imisebenzi enokwenziwa yinkqubo yokubona yomntu. Njengoqeqesho lwezenzululwazi, umbono wekhompyutha ujongene nethiyori esemva kweenkqubo ezenziweyo ezikhupha inkcazelo kwimifanekiso.	Inkundla eyizinhlobonhlobo egxile endleleni okungenziwa ngayo ukuba amakhompyutha athole ukuqonda kwezinga eliphezulu ngezithombe namavidiyo edijithali. Ngokombono wonjiniyela, ihlose ukwenza imisebenzi eyenziwa isistimu yokubuka komuntu. Ngokomthetho wesayensi, ukubona kwekhompyutha kukhathazeke ngendaba yamasistimu okwenziwa adonsa imininingwane ezithombeni.
google_research_ai_terminology_11	Conversational AI	gesprek-KI	I-AI yokuncokola	Ingxoxo yobuhlakani obakhiwe	A type of artificial intelligence (AI) that can simulate human conversation thanks to natural language processing (NLP) technologies.	’n Soort kunsmatige intelligensie (AI) wat menslike gesprekke kan naboots danksy tegnologie vir natuurlike taalverwerking (NTV).	Uhlobo lwe-artificial intelligence (AI) ekwazi ukulinganisa incoko yabantu ngoncedo lobuchwephesha bokulungiswa kolwimi lwendalo (NLP).	Uhlobo lobuchwepheshe bokwakhiwe (AI) olungalingisa ingxoxo yabantu ngenxa yobuchwepheshe bokucubungula kolimi lwemvelo (NLP).
google_research_ai_terminology_12	Data anonymization	data-anonimisering	Ukwenza idatha ifihlwe iinkcukacha	Ufihlomvelaphi lwemininingo	A process by which personal data is altered in such a way that a data subject can no longer be identified directly or indirectly, either by the data controller alone or in collaboration with any other party.	’n Proses waardeur persoonlike data só gewysig word dat ’n datasubjek nie meer direk of indirek deur óf die databeheerder alleen, óf in samewerking met enige ander party geïdentifiseer kan word nie.	Inkqubo apho idatha yomntu iguqulwa khona ngendlela yokuba umxholo wedatha ungasakwazi ukuchongwa ngokuthe ngqo okanye ngokungathanga ngqo, nokuba ngumlawuli wedatha yedwa okanye ngokubambisana nalo naliphi na elinye iqela.	Inqubo yokuguqulwa kwedatha yomuntu ngendlela yokuthi umnikazi wedatha angabe esakwazi ukutholwa ngokuqondile noma ngokungaqondile, kungaba ngesilawuli sedatha sisodwa noma sihlangene nanoma yikuphi okunye.
google_research_ai_terminology_13	Data augmentation	data-aanvulling	Ukwandiswa kwedatha	Isandisamininingo	A statistical technique which allows maximum likelihood estimation from incomplete data.	’n Statistiese tegniek wat ’n skatting van die grootste waarskynlikheid uit onvolledige data moontlik maak.	Ubuchule beenkcukacha-manani obuvumela uqikelelo lokunokwenzeka okukhulu okusuka kwidatha engaphelelanga.	Ikhono lezibalo elivumela ukulinganisa kokuqagela okukhulu ngedatha engaphelele.
google_research_ai_terminology_14	Data de-anonymization	datadeanonimisering	Ukwaziwa kwedatha yaziwe kwakhona	Ivezamvelaphi lemininingo	It is the practice of matching anonymous data (also known as de-identified data) with publicly available information, or auxiliary data, in order to discover the person the data belong to. Also known as 'data re-identification'.	Dit is die praktyk om anonieme data (ook bekend as gedeïdentifiseerde data) te pas by inligting wat publiek beskikbaar is, of aanvullende data, om uit te vind aan watter persoon die data behoort. Dit is ook bekend as “dataheridentifisering”.	Isenzo sokuthelekisa idatha engachazwanga (ekwabizwa ngokuba yidatha eyekisiweyo ukuchonga) enenkcukacha ezifumaneka esidlangalaleni, okanye iinkcukacha ezincedisayo, ukuze kufunyanwe ukuba ngubani umntu wale datha. Ikwabizwa ngokuba 'kukuphinda kuchongwe idatha'.	Inqubo yokuqhathanisa idatha engaziwa (eyaziwa nangokuthi yide-identified data) enemininingwane etholakala esidlangalaleni, noma idatha eseceleni, ukuze kutholakale umuntu ongumnikazi waleyo datha. Yaziwa nangokuthi 'yidata re-identification'.
google_research_ai_terminology_15	Data mining	dataontginning	Ukwembiwa kweenkcukacha	Ukuhlwaya imininingo	An interdisciplinary subfield of computer science and statistics with an overall goal of extracting information (with intelligent methods) from a data set and transforming the information into a comprehensible structure for further use.	’n Interdissiplinêre subvakgebied van rekenaarwetenskap en statistiek waarvan die oorkoepelende doel is om inligting (met intelligente metodes) uit ’n datastel te haal en die inligting tot ’n verstaanbare struktuur te omskep vir verdere gebruik.	Icandelo elingaphantsi kwezifundo zesayensi yekhompyutha neenkcukacha-manani enenjongo yokukhupha inkcazelo (ngeendlela ezikrelekrele) ukusuka kwiseti yedatha nokuguqula inkcazelo kwisakhiwo esiqondakalayo ukuze sisetyenziswe ngakumbi.	Inkundla encane eyizinhlobonhlobo yesayensi yekhompyutha nezibalo enomgomo wokudonsa imininingwane (ngezindlela ezihlakaniphile) kusethi yedatha nokuguqula leyo mininingwane ibe yisakhiwo esiqondakalayo ukuze isetshenziswe.
google_research_ai_terminology_16	Deep Learning	diepleer	Ukufunda nzulu	Ukufunda usebenzisa izici	Deep learning refers to a class of machine learning algorithms in which a hierarchy of layers is used to transform input data into a slightly more abstract and composite representation. For example, in an image recognition model, the raw input may be an image (represented as a tensor of pixels). The first representational layer may attempt to identify basic shapes such as lines and circles, the second layer may compose and encode arrangements of edges, the third layer may encode a nose and eyes, and the fourth layer may recognize that the image contains a face.	Diepleer verwys na ’n klas masjienleeralgoritmes waarin ’n hiërargie lae gebruik word om invoerdata tot ’n effens meer abstrakte en saamgestelde verteenwoordiging te omskep. Byvoorbeeld, in ’n prentherkenningmodel kan die onverwerkte invoer ’n prent wees (wat as ’n tensor pieksels verteenwoordig word). Die eerste verteenwoordigende laag kan probeer om basiese vorms, soos lyne en sirkels, te identifiseer; die tweede laag kan rangskikkings van rande saamstel en enkodeer; die derde laag kan neuse en oë enkodeer; en die vierde laag kan herken dat die prent ’n gesig bevat.	Ukufunda nzulu kubhekisele kwiklasi yokufunda koomatshini beealgorithm apho ukulandelelana kweeleya kusetyenziselwa ukuguqula idatha efakiweyo ibe yinto engabonakaliyo nomboniso odibeneyo. Umzekelo, kwimodeli yokuqaphela umfanekiso, umthombo wokufakwayo unokuba ngumfanekiso (omelwe njengetensor yeepikseli). Ileya yokuqala omeleyo unokuzama ukuchonga iimilo ezisisiseko ezifana nemigca kunye nezangqa, ileya yesibini inokuqamba ize idibanise ulungelelwaniso lwemiphetho, ileya yesithathu inokubhaqa impumlo namehlo, ize ileya yesine iqonde ukuba umfanekiso uqulathe ubuso.	Ideep learning ibhekisela esigabeni sama-algorithm okufunda komshini lapho amazinga ezendlalelo esetshenziselwa ukuguqula idatha efakwayo ibe yisimo esifihlekile nesididiyelwe. Ngokwesibonelo, kumodeli yokuqashelwa kwesithombe, okufakwayo okungahlungiwe kungase kube isithombe (esimelelwa yitensor yamaphikseli). Isendlalelo sokuqala esimelelayo singase sizame ukubona izinhlobo zezinto ezivamile njengolayini neziyingi, isendlalelo sesibili singase sibhale futhi sifake ikhodi ekuhlelweni kwamachopho, isendlalelo sesithathu singase sifake ikhodi yekhala namehlo, kanti isendlalelo sesine singase sibone ukuthi isithombe siqukethe ubuso.
google_research_ai_terminology_17	Deepfake	diepvervalsing	Ideepfake	Ifihlabuwuwa	Images, videos, or audio which are edited or generated using artificial intelligence tools, and which may depict real or non-existent people. They are a type of synthetic media.	Prente, video’s of oudio wat met gebruik van nutsgoed met kunsmatige intelligensie geredigeer of gegenereer is, en wat regte of niebestaande mense kan uitbeeld. Dit is ’n soort sintetiese media.	Imifanekiso, iividiyo, okanye okumanyelwayo okuedithiweyo okanye okuveliswa kusetyenziswa izixhobo zobukrelekrele ezenziweyo, ezinokuthi zimele abantu bokwenyani okanye abangekhoyo. Ziluhlobo lwemidiya eyenziweyo.	Izithombe, amavidiyo, noma okulalelwayo okuhlelwa noma kukhiqizwe ngamathuluzi wobuchwepheshe bokwakhiwe, futhi okungabonisa abantu bangempela noma abangekho ngokoqobo. Kuyizinhlobo zesynthetic media.
google_research_ai_terminology_18	Explainable AI	verklaarbare KI	I-AI ecacisekayo	Ubuhlakani Obakhiwe Obuchazekayo	XAI / Either refers to an artificial intelligence (AI) system over which it is possible for humans to retain intellectual oversight, or refers to the methods to achieve this.	VKI / Dit verwys óf na ’n stelsel met kunsmatige intelligensie (AI) waaroor dit vir mense moontlik is om intellektuele toesig te behou, óf dit verwys na die metodes om dit reg te kry.	XAI / Ibhekisa kwinkqubo yobukrelekrele bokwenziwa (AI) apho kunokwenzeka ukuba abantu bagcine ukongamela ubukrelekrele, okanye ibhekisa kwiindlela zokwazi ukuphumeza oku.	Ibhekisela kusistimu yobuchwepheshe bokwakhiwe (AI) eyenza kwenzeke ukuthi abantu baqhubeke beqondisa ngokuhlakanipha, noma kubhekisela ezindleleni zokukufinyelela lokhu.
google_research_ai_terminology_19	Fairness	regverdigheid	Ubulungisa	Ukungachemi	in AI / Fairness in machine learning refers to the various attempts at correcting algorithmic bias in automated decision processes based on machine learning models. Decisions made by computers after a machine-learning process may be considered unfair if they were based on variables considered sensitive. For example gender, ethnicity, sexual orientation or disability.	in KI / Regverdigheid in masjienleer verwys na die verskillende pogings om algoritmiese vooroordeel in geoutomatiseerde besluitnemingprosesse op grond van masjienleermodelle reg te stel. Besluite wat ná ’n masjienleerproses deur rekenaars geneem word, kan as onregverdig beskou word as hulle gegrond is op veranderlikes wat as sensitief beskou word. Byvoorbeeld gender, etnisiteit, seksuele oriëntasie of gestremdheid.	kwi-AI / I-Fairness ekufundweni koomatshini ibhekiselele kwiinzame ezahlukeneyo zokulungisa ukuthatha icala kweealgorithm kwiinkqubo zezigqibo ezizenzekelayo ezisekelwe kwiimodeli zokufundwa koomatshini. Izigqibo ezenziwe ziikhompyutha emva kwenkqubo yokufunda koomatshini zinokujongwa njengezingafanelekanga ukuba bezisekelwe kwizinto ezithathwa njengezinovakalelo. Umzekelo isini, ubuhlanga, ukuziqhelanisa ngesondo okanye ukukhubazeka.	ku-AI / Ukungachemi ekufundeni komshini kubhekisela emizamweni ehlukahlukene yokulungisa ukuchema kwama-algorithm ezinqubweni zezinqumo ezizenzekelayo ezisekelwe kumamodeli okufunda komshini. Izinqumo ezenziwa amakhompyutha ngemva kwenqubo yokufunda komshini zingase zibhekwe njengezingafanele uma zisekelwe ezingxenyekazini ezibhekwa njengezizwelayo. Ngokwesibonelo ubulili, ubuzwe, ubulili oyamaniseka nabo noma ukukhubazeka.
google_research_ai_terminology_20	Few-shot prompting	veelvoorbeeldinstruksieteks	Uyalelo olumizekelo mininzi	Ukuyalela ngokumbalwa	A prompting strategy that includes two or more examples of the desired input and output	’n Porboodskapstrategie wat twee of meer voorbeelde van die verlangde invoer en uitvoer insluit	Isicwangciso esiyalelayo esibandakanya imizekelo emibini okanye ngaphezulu yokufakwayo okufunekayo nesiphumo	Isu leprompting oluhlanganisa izibonelo ezimbili noma ngaphezulu zokufakwayo nomphumela ofunwayo
google_research_ai_terminology_21	Fine-tuning	verfyning	Ukutyhuna kakuhle	Ukucolisisa	In deep learning, fine-tuning is an approach to transfer learning in which the parameters of a pre-trained model are trained on new data.	In diepleer is fyn instelling ’n benadering om leer oor te dra waarin die parameters van ’n voorafopgeleide model met nuwe data opgelei word.	Ekufundeni nzulu, i-fine-tuning yindlela yokudlulisa ukufunda apho iipharamitha zemodeli eqeqeshwe kwangaphambili ziqeqeshelwe idatha entsha.	Ekufundeni ngokujulile, ifine-tuning iyindlela yokudlulisela ukufunda lapho amapharamitha wemodeli eqeqeshwe kusengaphambili eqeqeshelwa idatha entsha.
google_research_ai_terminology_22	Generative AI	generatiewe KI	I-AI yokuvelisa	Ubuhlakani Obakhiwe Obukhiqizayo	Artificial intelligence capable of generating text, images, videos, or other data using generative models, often in response to prompts. Generative AI models learn the patterns and structure of their input training data and then generate new data that has similar characteristics.	Kunsmatige intelligensie wat die vermoë het om teks, prente, video’s of ander data met gebruik van generatiewe modelle te genereer, dikwels in reaksie op porboodskappe. Generatiewe-AI-modelle leer die patrone en struktuur van hul invoeropleidingdata en genereer dan nuwe data wat soortgelyke kenmerke het.	I-artificial intelligence ekwazi ukuvelisa umbhalo, imifanekiso, iividiyo, okanye enye idatha isebenzisa imodeli yokuzivelisela, ehlala iphendula imiyalelo. Iimodeli ze-AI Yokuvelisa zifunda iiphatheni nolwakhiwo lwedatha yoqeqesho lwazo zize emva koko zivelise idatha entsha eneempawu ezifanayo.	Ubuchwepheshe bokwakhiwe bukwazi ukukhiqiza umbhalo, izithombe, amavidiyo, nenye idatha ngamamodeli akhiqizayo, ngokuvamile ngokusabela emiyalelweni. Amamodeli we-AI ekhiqizayo afunda amaphethini nesakhiwo sedatha yokuqeqeshwa yokufakwayo bese ekhiqiza idatha entsha enezici ezifanayo.
google_research_ai_terminology_23	Hallucination	hallusinasie	Ukubona izinto ezingekhoyo	Ukudideka	A response generated by AI which contains false or misleading information presented as fact.	’n Antwoord gegenereer deur AI wat vals of misleidende inligting bevat, maar as ’n feit voorgehou word.	Impendulo eveliswa yi-AI equlethe inkcazelo engeyonyani okanye elahlekisayo eboniswe njengenyaniso.	Impendulo ekhiqizwa yi-AI equkethe imininingwane engamanga noma edukisayo enikezwa njengeyiqiniso.
google_research_ai_terminology_24	Human-in-the-Loop	mens-wat-ingesluit-word	Ihuman-in-the-Loop	Inxumano Yomuntu Nomshini	HITL / In machine learning, HITL is used in the sense of humans aiding the computer in making the correct decisions in building a model.	MWIW / In masjienleer word MWIW gebruik in die sin van mense wat die rekenaar bystaan om die korrekte besluite te neem wanneer ’n model gebou word.	HITL / Ekufundeni koomatshini, i-HITL isetyenziswa kwingqiqo yabantu encedisa ikhompyutha ekwenzeni izigqibo ezichanekileyo ekwakheni imodeli.	Ekufundeni komshini, i-HITL isetshenziswa ngomqondo wokuthi abantu basiza ikhompyutha ekwenzeni izinqumo ezinembile ekwakheni imodeli.
google_research_ai_terminology_25	Large Language Model	groot taalmodel	IModeli yoLwimi eNkulu	Insuselakuyo Yolimi Ngokubanzi	LLM / A computational model capable of language generation or other natural language processing tasks. As language models, LLMs acquire these abilities by learning statistical relationships from vast amounts of text during a self-supervised and semi-supervised training process.	GTM / ’n Rekenaarmodel wat taal kan genereer of ander take met natuurlike taalverwerking kan uitvoer. As taalmodelle verkry GTM’e hierdie vermoëns deur statistiese verhoudings uit ontsaglike hoeveelhede teks te leer tydens ’n opleidingproses onder selftoesig of semitoesig.	LLM / Imodeli yekhompyutha ekwaziyo ukuvelisa ulwimi okanye eminye imisebenzi yokulungisa ulwimi lwendalo. Njengeemodeli zolwimi, ii-LLM zifumana obu buchule ngokufunda unxulumano lwamanani ukusuka kumthamo omkhulu wezibhalo ngexesha lenkqubo yoqeqesho oluzigadelayo noluphantsi kweliso eliphakathi.	Imodeli yekhompyutha ekwazi ukukhiqiza ulimi noma eminye imisebenzi yokucubungula kolimi lwemvelo. Njengamamodeli wolimi, ama-LLM athola lawa makhono ngokufunda ubuhlobo bezibalo emananini amaningi ombhalo phakathi nenqubo yokuqeqesha ozigada ngokwakho kuyo negadwa ngokwengxenye.
google_research_ai_terminology_26	Machine Learning	masjienleer	IMachine Learning	Ukufunda Komshini	ML / A field of study in artificial intelligence concerned with the development and study of statistical algorithms that can learn from data and generalize to unseen data and thus perform tasks without explicit instructions.	ML / ’n Studierigting in kunsmatige intelligensie wat ingestel is op die ontwikkeling en bestudering van statistiese algoritmes wat uit data kan leer en dit vir onbekende data kan veralgemeen, en gevolglik take sonder uitdruklike instruksies kan uitvoer.	ML / Icandelo lokufunda kwi-artificial intelligence ejongene nophuhliso nokufunda ii-algorithim zamanani anokuthi afunde kwidatha kwaye enze ngokubanzi kwidatha engabonakaliyo kwaye ngaloo ndlela enze imisebenzi ngaphandle kwemiyalelo ecacileyo.	Inkundla yokucwaninga ebuchwephesheni bokwakhiwe egxile ekwakhekeni nasocwaningweni lwama-algorithm ezibalo engafunda kudatha futhi iyiguqulele kudatha engabonakali ngaleyo ndlela yenze imisebenzi ngaphandle kweziqondiso ezicacile.
google_research_ai_terminology_27	Multimodal learning	multimodale leer	Ukufunda ngeeModali ezininzi	Ukufunda ngezinsuselakuyo eziningi	A type of deep learning using multiple modalities of data, such as text, audio, or images.	’n Soort diepleer wat veelvuldige modaliteite data, soos teks, oudio of prente, gebruik.	Uhlobo lokufunda olunzulu usebenzisa iindela ezininzi zedatha, ezinjengemibhalo, okumanyelwayo, okanye imifanekiso.	Uhlobo lokufunda okujulile olusebenzisa amamodali wedatha angaphezu kweyodwa, njengombhalo, okulalelwayo, noma izithombe.
google_research_ai_terminology_28	Natural Language Processing	natuurliketaalverwerking	UkuSetyenzwa koLwimi lweNdalo	Ukucubungula Kolimi Lwemvelo	NLP / An interdisciplinary subfield of computer science and artificial intelligence, primarily concerned with providing computers with the ability to process data encoded in natural language. Major tasks in natural language processing are speech recognition, text classification, natural-language understanding, and natural-language generation.	NTV / ’n Interdissiplinêre subvakgebied in rekenaarwetenskap en kunsmatige intelligensie wat hoofsaaklik daarop ingestel is om aan rekenaars die vermoë te gee om data te verwerk wat in natuurlike taal geënkodeer is. Belangrike take in natuurlike taalverwerking is spraakherkenning, teksklassifikasie, natuurlike taalbegrip en natuurlike taalgenerering.	NLP / Indawo engaphantsi kwezifundo zenzululwazi yekhompyutha ne-artificial intelligence, yeyona nto ijongene nokubonelela iikhompyutha ngesakhono sokucubungula idatha efakwe ngekhowudi ngolwimi lwendalo. Imisebenzi ephambili ekusetyenzweni kolwimi lwendalo kukunakanwa kwentetho, ukuhlelwa kwemibhalo, ukuqonda ulwimi lwendalo, nokuveliswa kolwimi lwendalo.	NLP / Inkundla engaphezu kweyodwa yesayensi yekhompyutha nobuchwepheshe bokwakhiwe, egxile ngokuyinhloko ekunikezeni amakhompyutha ikhono lokucubungula idatha ebhalwe ngekhodi yolimi olungokwemvelo. Imisebenzi emikhulu ekucubunguleni kolimi lwemvelo ukubonwa kwengxoxo, ukuhlunga umbhalo, ukuqonda ulimi lwemvelo, nokukhiqiza ulimi lwemvelo.
google_research_ai_terminology_29	Noisy data	raserige data	Idatha eninzi engelo ncedo	Imininingo enolwazi olungadingeki	Data with a large amount of additional meaningless information in it called noise.	Data met ’n groot hoeveelheid bykomende, betekenislose inligting daarin wat geraas genoem word.	Idatha enesixa esikhulu senkcazelo eyongezelelweyo engenantsingiselo kuyo ibizwa ngokuba yi-noise.	Idatha enenani elikhulu lemininingwane engenancazelo kuyo ebizwa ngomsindo.
google_research_ai_terminology_30	One-shot prompting	eenvoorbeeldinstruksieteks	Uyalelo olumzekelo mnye	Ukuyalelela ngokukodwa	A prompting strategy that includes one example of the desired input and output	’n Porboodskapstrategie wat een voorbeeld van die verlangde invoer en uitvoer insluit	Isicwangciso sokuyalela esibandakanya umzekelo omnye wento efakwayo efunekayo nesiphumo	Isu lomyalelo elibandakanya isibonelo esisodwa sokufakwayo nomphumela ofiselekayo
google_research_ai_terminology_31	Open source	oopbron	Umthombo ovulelekileyo	Umthombo ovulekile	A source code (a plain text computer program written in a programming language) that is made freely available for possible modification and redistribution.	’n Bronkode (’n rekenaarprogram in skoonteks wat in ’n programmeertaal geskryf is) wat vrylik beskikbaar gestel word vir moontlike wysiging en herverspreiding.	Ikhowudi yomthombo (inkqubo yekhompyutha ebhaliweyo ecacileyo ebhalwe ngolwimi lwenkqubo) eyenziwe ifumaneke ngokukhululekileyo ukuze ikwazi ukuguqulwa nokusasazwa kwakhona.	Ikhodi yomthombo (uhlelo lwekhompyutha lombhalo ongenalutho olubhalwe ngolimi lweprogramming) etholakala mahhala ukuze iguqulwe futhi isakazwe kabusha.
google_research_ai_terminology_32	Output	afvoer	Imveliso	Umkhiqizo	The information or creative work that an AI tool produces after it is prompted, such as an answer to a question, a text summary, an image, or a piece of music	Die inligting of kreatiewe werk wat ’n AI-nutsding lewer nadat dit gepor is, soos ’n antwoord op ’n vraag, ’n teksopsomming, ’n prent of ’n stuk musiek	Inkcazelo okanye umsebenzi wokudala oveliswa sisixhobo se-AI emva kokuyalelwa, njengempendulo yombuzo, isishwankathelo sombhalo, umfanekiso, okanye isiqwenga somculo	Imininingwane noma umsebenzi wobuciko okhiqizwa yithuluzi le-AI ngemva komyalelo, njengempendulo embuzweni, isifinyezo sombhalo, isithombe, noma umculo
google_research_ai_terminology_33	Prompt	instruksieteks	Umyalelo	Umyalelo	An instruction, usually expressed in natural language, used to get the desired output from a generative AI engine.	’n Instruksie, gewoonlik uitgedruk in natuurlike taal, wat gebruik word om die verlangde uitvoer van ’n generatiewe-AI-enjin te kry.	Umyalelo, odla ngokuchazwa ngolwimi lwendalo, osetyenziselwa ukufumana imveliso efunekayo kwi-injini ye-AI yokuvelisa.	Isiqondiso, ngokuvamile esivezwa ngolimi olungokwemvelo, esisetshenziselwa ukuthola umphumela ofiswayo enjinini ye-AI ekhiqizayo.
google_research_ai_terminology_34	Prompt engineering	instruksieteksontwerp	Ubunjinelimyalelo	Ubunjiniyela bemiyalelo	The process of structuring an instruction that can be interpreted and understood by a generative AI model.	Die proses om ’n instruksie te struktureer wat deur ’n generatiewe-AI-model geïnterpreteer en verstaan kan word.	Inkqubo yokumisela umyalelo onokutolikwa uze uqondwe yimodeli ye-AI yokuvelisa.	Inqubo yokwakha umyalelo ongase uchazwe futhi uqondwe yimodeli ye-AI ekhiqizayo.
google_research_ai_terminology_35	Pseudonymization	pseudonimisering	IPseudonymization	Ukuzivezambumbulu	The process of obscuring data with the ability to re-identify it later. Personally identifiable information fields within a data record are replaced by one or more artificial identifiers, or pseudonyms.	Die proses om data te verdoesel met die vermoë om dit later te heridentifiseer. Velde met persoonlik identifiseerbare inligting in ’n datarekord word vervang met een of meer kunsmatige identifiseerders, of pseudonieme.	Inkqubo yokufihla idatha nokwazi ukuyichonga kwakhona kamva. Indawo yenkcazelo echongiweyo ngokobuqu kwirekhodi yedatha ethathelwe indawo sisifanisi esinye, okanye ngapezulu, okanye ii-pseudonym.	Inqubo yokufihla idatha ngendlela yokukwazi ukuyibona futhi kamuva. Izinkundla zolwazi olukhomba umuntu siqu (PII) erekhodini ledatha ziyakhishwa kufakwe izinkomba zokwenziwa ezizodwa noma ngaphezulu, noma amapseudonyms.
google_research_ai_terminology_36	Raw data	onverwerkte data	Idatha ekrwada	Imininingo engahlungiwe	Also known as primary data, are data (e.g., numbers, instrument readings, figures, etc.) collected from a source, prior to cleaning or analysis.	Ook bekend as primêre data – dit is data (byvoorbeeld nommers, instrumentlesings, syfers, ensovoorts) wat van ’n bron af ingesamel is, voordat dit opgeruim of ontleed is.	Ekwabizwa ngokuba yidatha yangaphambili, yidatha (umzekelo, amanani, ukufundwa kwezixhobo, amanani, njalo-njalo.) eqokelelwe kumthombo ngaphambi kokucoca okanye uhlalutyo.	Yaziwa nangokuthi idatha eyinhloko, iyidatha (isib., izinombolo, ukufundwa kwethuluzi, izinombolo, njll.) eqongelelwa emthonjeni, ngaphambi kokuhlanza noma ukuhlaziya.
google_research_ai_terminology_37	Reinforcement Learning	versterkingleer	Ukufunda ngokuKhuthazwa	Ukufunda Ukuthatha Izinqumo	RL / An interdisciplinary area of machine learning and optimal control concerned with how an intelligent agent ought to take actions in a dynamic environment in order to maximize the cumulative reward.	VL / ’n Interdissiplinêre terrein van masjienleer en optimale beheer wat gemoeid is met hoe ’n intelligente agent in ’n dinamiese omgewing behoort op te tree om die kumulatiewe beloning so groot moontlik te maak.	RL / Icandelo lokufunda koomatshini nolawulo olufanelekileyo olunenkxalabo ngendlela iarhente ekrelekrele ekufuneka ithathe ngayo amanyathelo kwindawo eguqukayo ukuze kwandiswe umvuzo owongezelekayo.	Indawo engaphezu kweyodwa yokufunda komshini nokulawula okuthuthukisiwe okugxile endleleni i-ejenti ehlakaniphile okufanele ithathe ngayo izinyathelo endaweni eshintshashintshayo ukuze kuthuthukiswe umvuzo oqongelelwe.
google_research_ai_terminology_38	Robotics	robotika	Irobhothiksi	IRobotiksi	The interdisciplinary study and practice of the design, construction, operation, and use of robots (a machine—especially one programmable by a computer—capable of carrying out a complex series of actions automatically).	Die interdissiplinêre bestudering en toepassing van die ontwerp, konstruksie, bedryf en gebruik van robotte (’n masjien – veral een wat deur ’n rekenaar geprogrammeer kan word – wat in staat is om ’n komplekse reeks handelinge outomaties uit te voer).	Uphononongo lwezifundo ezahlukahlukeneyo nokusebenza koyilo, ukwakhiwa, ukusebenza, kunye nokusetyenziswa kweerobhothi (umatshini–ngakumbi omnye ocwangciswayo yikhompyutha–ekwaziyo ukwenza uluhlu olunzima lwezenzo ngokuzenzekelayo).	Ucwaningo olungaphezu kolulodwa nomsebenzi wedizayini, ukwakha, ukusebenza, nokusetshenziswa kwamarobhothi (umshini—ikakhulu ongahlelwa yikhompyutha—okwazi ukufeza uchungechunge lwezenzo eziyinkimbinkimbi ngokuzenzekelayo).
google_research_ai_terminology_39	Speech recognition	spraakherkenning	Ukunakanwa kwentetho	Inzwagezwi	An interdisciplinary subfield of computer science and computational linguistics that develops methodologies and technologies that enable the recognition and translation of spoken language into text by computers.	’n Interdissiplinêre subvakgebied van rekenaarwetenskap en rekenaarlinguistiek wat metodologieë en tegnologieë ontwikkel wat die herkenning en vertaling van gesproke taal in teks deur rekenaars moontlik maak.	Icandelo elingaphantsi kwezifundo ezahlukeneyo zenzululwazi yekhompyutha neelwimi zekhompyutha eziphuhlisa iindlela nobuchwepheshe obuvumela ukuqondwa nokuguqulelwa kolwimi oluthethwayo kwisicatshulwa ngeekhompyutha.	Inkundla engaphezu kweyodwa yesayensi yekhompyutha nolimi lwekhompyutha olusungula izindlela nobuchwepheshe obuvumela ukuqashelwa nokuhunyushwa kolimi olukhulunywayo ngombhalo emakhompyutheni.
google_research_ai_terminology_40	Supervised learning	leer onder toesig	Ukufunda okujongiweyo	Ukufunda okuqashelwe	A paradigm in machine learning where input objects and a desired output value train a model.	’n Paradigma in masjienleer waar invoervoorwerpe en ’n verlangde uitvoerwaarde ’n model oplei.	I-paradigm ekufundeni koomatshini apho izinto ezifakwayo nexabiso elifunekayo lemveliso liqeqesha imodeli.	Isibonelo sokufunda komshini lapho izinto ezifakwayo nenani lomphumela ofiswayo ziqeqesha imodeli.
google_research_ai_terminology_41	Synthetic media	sintetiese media	Imidiya yokwenziwayo	Ezokwazisa ezikhiqizwa umshini	A catch-all term for the artificial production, manipulation, and modification of data and media by automated means, especially through the use of artificial intelligence algorithms, such as for the purpose of misleading people or changing an original meaning. Also known as AI-generated media.	’n Versamelterm vir die kunsmatige produksie, manipulasie en modifikasie van data en media deur geoutomatiseerde metodes, veral deur die gebruik van algoritmes met kunsmatige intelligensie, soos met die doel om mense te mislei of om ’n oorspronklike betekenis te verander. Ook bekend as AI-gegenereerde media.	Intetho ehlala isetyenziswa yemveliso eyenziweyo , ukwenziwa nokuguqulwa kwedatha kunye nemidiya ngeendlela ezizenzekelayo, ngokukodwa ngokusetyenziswa kwee-algorithms ze-artificial intelligence, njengenjongo yokulahlekisa abantu okanye ukutshintsha intsingiselo yangaphambili. Ekwabizwa ngokuba yimidiya eveliswe yi-AI.	Itemu elihlanganisa konke lomkhiqizo owakhiwe, ukufuzisela, nokuguqulwa kwedatha nemidiya ngokuzenzakalelayo, ikakhulu ngokusebenzisa ama-algorithms wobuchwepheshe bokwakhiwe, ngenjongo yokudukisa abantu noma ukushintsha incazelo yasekuqaleni. Yaziwa nangokuthi yimidiya ekhiqizwe yi-AI.
google_research_ai_terminology_42	Transfer learning	oordragleer	Udluliselo lokufundiweyo	Ukufunda kokudlulisela	Transfer learning (TL) is a technique in machine learning (ML) in which knowledge learned from a task is re-used in order to boost performance on a related task.[1] For example, for image classification, knowledge gained while learning to recognize cars could be applied when trying to recognize trucks.	Oordragleer (OL) is ’n tegniek in masjienleer (ML) waarin kennis wat uit ’n taak geleer is, weer gebruik word om die prestasie met ’n verwante taak te verbeter.[1] Byvoorbeeld, vir prentklassifikasie kan die kennis wat verkry is terwyl daar geleer is om motors te herken, toegepas word wanneer daar geprobeer word om vragmotors te herken.	Ukufunda okudluliswayo (TL) bubuchule ekufundeni koomatshini (ML) apho ulwazi olufundwe kumsebenzi luphinda lusetyenziswe ukuze kunyuswe ukusebenza komsebenzi onxulumeneyo.[1] Umzekelo, kuhlelo lwemifanekiso, ulwazi olufunyenwe ngelixa ufunda ukuqaphela iimoto lunokusetyenziswa xa kuzanywa ukuqatshelwa iilori.	Ukufunda kokudlulisela (TL) ingubuciko ekufundeni komshini (ML) lapho ulwazi olutholwe emsebenzini lusetshenziswa kabusha ukuze kuthuthukiswe ukusebenza komsebenzi ohlobene.[1] Ngokwesibonelo, ekuhlungeni isithombe, ulwazi olutholwe ngesikhathi kufundwa ukuqaphela izimoto lungasebenza uma kuzanywa ukuqaphela amaloli.
google_research_ai_terminology_43	Unsupervised learning	leer sonder toesig	Ukufunda okungajongwangwa	Ukufunda okungagadiwe	A framework in machine learning where, in contrast to supervised learning, algorithms learn patterns exclusively from unlabeled data.	’n Raamwerk in masjienleer waar – anders as met leer onder toesig – algoritmes patrone eksklusief uit ongemerkte data leer.	Isakhelo ekufundweni koomatshini apho, ngokuchaseneyo nokufunda okugadiweyo, ii-algorithm zifunda iipateni ngokukodwa kwiidatha ezingabhalwanga.	Ifremuwekhi ekufundeni komshini lapho, ngokungafani nokufunda okugadiwe, ama-algorithms afunda amaphethini ngokukhethekile kudatha engenalebula.
google_research_ai_terminology_44	Zero-shot prompting	geenvoorbeeldinstruksieteks	Uyalelo olungena mizekelo	Ukuyalela ngeze	A prompting strategy that doesn’t include any examples about the desired input and output	’n Porboodskapstrategie wat geen voorbeelde van die verlangde invoer en uitvoer insluit nie	Isicwangiso sokuyalela esingabandakanyi nayiphi na imizekelo ngokufakwayo okufunekayo nesiphumo.	Isu lokukhumbuza elingabandakanyi noma iziphi izibonelo ezimayelana nokufakwayo kanye nomphumela ofiswayo
google_research_ai_terminology_45	A/B testing	A/B-toetsing	Uvavanyo lweA/B	Ukuhlolakuqhathanisa imikhiqizo	A statistical way of comparing two (or more) techniques—the A and the B. Typically, the A is an existing technique, and the B is a new technique. A/B testing not only determines which technique performs better but also whether the difference is statistically significant.	’n Statistiese manier om twee (of meer) tegnieke te vergelyk - die A en die B. Gewoonlik is die A ’n bestaande tegniek en die B ’n nuwe tegniek. A/B-toetse bepaal nie net watter tegniek beter presteer nie, maar ook of die verskil wel statisties relevant is.	Indlela yezibalo yokuthelekisa iindlela ezimbini (okanye ngaphezulu)-i-A kunye ne-B. Ngokuqhelekileyo, i-A yinkqubo ekhoyo, kwaye i-B yindlela entsha. Uvavanyo lwe-A/B alubonisi nje kuphela ukuba boluphi na ubugcisa obusebenza ngcono kodwa nokuba umahluko ubalulekile ngokwezibalo.	Indlela yezibalo yokuqhathanisa izindlela - u-A kanye no-B. Ngokujwayelekile, u-A uyindlela ekhona, bese u-B abe yindlela entsha. Ukuhlola kwe-A/B akugcini ngokunquma ukuthi iyiphi indlela esebenza kangcono kodwa nokuthi umehluko wezibalo ubalulekile na.
google_research_ai_terminology_46	Ablation	ablasie	Ukukhutshwa	ukukhipha izakhi	In artificial intelligence (AI), particularly machine learning (ML), ablation is the removal of a component of an AI system. An ablation study investigates the performance of an AI system by removing certain components to understand the contribution of the component to the overall system.	In kunsmatige intelligensie (KI), spesifiek masjienleer (ML), is ablasie die verwydering van ’n komponent van ’n KI-stelsel. ’n Ablasiestudie ondersoek die werkverrigting van ’n KI-stelsel deur sekere komponente te verwyder om die bydrae van die komponent tot die algehele stelsel te verstaan.	Kubukrelekrele obenziweyo (iAI), ngokukodwa ukufundwa komatshini (ML), ukukhutshwa kukukhutshwa kwecandelo lenkqubo yeAI. Uphononongo lokuchithwa luphanda ukusebenza kwenkqubo yeAI ngokususa amacandelo athile ukuqonda igalelo lecandelo kwinkqubo yonke.	Ubuchwepheshe obusebenza ngokungasikho okwemvelo (i-AI), ikakhulukazi ukufunda ngomshini (i-ML), ukukhipha izakhi zohlelo lwe-AI. Ucwaningo lwezakhi luhlola ukusebenza kohlelo lwe-AI ngokususa izakhi ezithile ukuze uqonde igalelo lezakhi ohlelweni lulonke.
google_research_ai_terminology_47	AI Adoption	KI-ingebruikneming	Uguqulelo kwiAI	Ukwamukela Ubuhlakani Obakhiwe	In computing, adoption means the transfer (conversion) between an old system and a target system in an organization (or more broadly, by anyone). In regard to AI, it means using artificial intelligence to improve workflows and decision-making processes	In rekenaarwese beteken ingebruikneming die oordrag (omskakeling) tussen ’n ou stelsel en ’n teikenstelsel in ’n organisasie (of wyer nog, deur enigiemand). Met betrekking tot KI beteken dit dat kunsmatige intelligensie gebruik word om werkvloei en besluitnemingprosesse te verbeter.	Ekusebenzeni kweekhompyutha, uguqulelo luthetha ukudluliselwa (ukuguqulwa) phakathi kwenkqubo endala nenkqubo eyithagethi kwintlangano (okanye ngokubanzi, nguye nabani na). Ngokuphathelele iAI, kuthetha ukusebenzisa ubukrelekrele obenziweyo ukuze kuphuculwe ukuhamba komsebenzi neenkqubo zokwenza izigqibo.	Kwezocwaningo lwamakhompyutha, ukuguqukela kusho ukudlulisa (ukushintsha) usuke ohlelweni oludala kanye nohlelo okuphokophelwe kulo esikhungweni (noma kabanzi, okwenziwa yinoma wubani). Maqondana ne-AI, kusho ukusebenzisa ubuchwephe obusebenza ngokungasikho okwemvelo ukuthuthukisa umsebenzi kanye nezinqubo zokuthatha izinqumo
google_research_ai_terminology_48	Anomaly detection	afwykingbespeuring	Ufunyaniso lokungaqhelekanga	Ukuhlonza okungavamile	In machine learning, this is the process of using machine learning models to identify anomalies rapidly. This serves several purposes, whether to maintain clean, high-quality data that you will use for processing or specific business purposes.	In masjienleer is dit die proses waardeur masjienleermodelle gebruik word om onreëlmatighede vinnig te identifiseer. Dit dien verskeie doele, onder andere die behoud van skoon, hoëgehaltedata wat gebruik word vir prosessering of spesifieke besigheidsdoeleindes.	Ukufunda koomatshini, le yinkqubo yokusebenzisa iimodeli zokufunda ngoomatshini ukuchonga izinto ezingaqhelekanga ngokukhawuleza. Oku kusebenzela iinjongo ezininzi, nokuba kukugcina idatha ecocekileyo, ekumgangatho ophezulu oya kuyisebenzisela ukuprosesa okanye iinjongo ezithile zoshishino.	Ekufundeni ngemishini, lokhu kuyinqubo yokusebenzisa izinhlelo zokufunda ngemishini sokuhlonza okungavamile ngokushesha. Lokhu kusebenza ngokwezinhloso ezihlukene, okungaba wukugcina imininingwane engenasici, eseqophelweni eliphezulu ozoyisebenzisela ukulungisa noma izinhloso zebhizinisi ezibaluliwe.
google_research_ai_terminology_49	Artifical general intelligence	kunsmatige algemene intelligensie	Ubukrelekrele obenziweyo ngokubanzi	Ubuhlakani Obakhiwe Obuvamile	A non-human mechanism that demonstrates a broad range of problem solving, creativity, and adaptability. For example, a program demonstrating artificial general intelligence could translate text, compose symphonies, and excel at games that have not yet been invented.	’n Niemenslike meganisme wat ’n algemene, wye reeks probleemoplossing, kreatiwiteit en aanpasbaarheid demonstreer. Byvoorbeeld, ’n program wat kunsmatige intelligensie demonstreer kan teks vertaal, simfonieë komponeer en uitblink in speletjies wat nog nie geskep is nie.	Indlela engeyoyamntu ebonisa uluhlu olubanzi lokusombulula iingxaki, ukuyila, nokuguquguquka. Umzekelo, iprogramu ebonisa ubukrelekrele obenziweyo ngokubanzi bunokuguqulela umbhalo, buqambe iisimfoni, buze bugqwese kwimidlalo engekaveliswa.	Inqubo engeziwa abantu yokuveza uhla olubanzi lokuxazulula inkinga, ubuciko, kanye nokuhambisana nezimo. Isibonelo, uhlelo olubonisa ubuchwepheshe obusebenza ngokungasikho okwemvelo jikelele obungakwazi ukuhumusha umbhalo, buqambe imiculo, futhi buhambe phambili emidlalweni engakasungulwa.
google_research_ai_terminology_50	Artificial Narrow Intelligence	kunsmatige beperkte intelligensie	Ubukrelekrele oBenziweyo oBumxinwa	Ubuhlakani Obakhiwe Obubhuncene	ANI / Narrow AI can be classified as being “limited to a single, narrowly defined task. Most modern AI systems would be classified in this category.” Some examples of narrow AI are self-driving cars and robot systems used in the medical field.	KBI / Beperkte KI kan geklassifiseer word as "beperk tot ’n enkel, noukeurig gedefinieerde taak". Meeste moderne KI-stelsels sal in hierdie kategorie geklassifiseer word". ’n Paar voorbeelde van beperkte KI is selfbestuurkarre en robotstelsels wat in die mediese veld gebruik word.	ANI / I-AI Emxinwa inokuhlelwa “njengesikelwe umda kumsebenzi omnye ochaziweyo. Uninzi lweenkqubo ze-AI zanamhlanje zingahlelwa kolu didi". Eminye imizekelo ye-AI emxinwa ziimoto eziziqhubayo neenkqubo zerobhothi ezisetyenziswa kwicandelo lezonyango.	I-AI encinyanane okungathiwa "inomkhawulo wokwenza umsebenzi owodwa umcinyane ochaziwe. Izinhlelo ze-AI zesimanje eziningi zingafakwa kulesi sigaba." Ezinye izibonelo zobuchwepheshe obusebenza ngokungasikho okwemvelo obuncinyane yizimoto ezimoto ezizishayelayo kanye nezinhlelo zamarobhothi ezisetshenziswa emkhakheni wezokwelapha.
google_research_ai_terminology_51	AutoML	OutoML	iAutoML	Ukufunda Kwemishini Ngokuzenzekela	Automated Machine Learning / Automated machine learning (AutoML) is the process of automating the tasks of applying machine learning to real-world problems. It is the combination of automation and ML.	outomatiese masjienleer / Outomatiese masjienleer (OutoML) is die proses waardeur die take waarin masjienleer op regtewêreldprobleme toegepas word, geoutomatiseer word. Dis die kombinasie van outomatisering en masjienleer.	ukuFunda ngoMatshini okuZenzekelayo / Ukufunda ngomatshini ngokuzenzekelayo (iAutoML) yinkqubo yokuzenzela imisebenzi yokufaka umatshini wokufunda kwiingxaki zehlabathi lokwenyani. Yindibaniselwano yokuzenzela neML.	Ufunda kwemishini ngokuzenzekela (i-AutoML) yinqubo yokuhlela ukuba imisebenzi izenzekele usebenzisa ukufunda kwemishini ezinkingeni zasemhlabeni zangempela. Kuhlanganisa ukuzenzekela kanye ne-ML.
google_research_ai_terminology_52	Big data	groot data	Idatha enkulu	Imininingo emikhulu	Primarily refers to data sets that are too large or complex to be dealt with by traditional data-processing application software. Big data usually includes data sets with sizes beyond the ability of commonly used software tools to capture, curate, manage, and process data within a tolerable elapsed time.	Verwys hoofsaaklik na datastelle wat te groot of kompleks is om deur tradisionele appsagteware vir dataprosessering gehanteer te word. Grootdata sluit gewoonlik datastelle in met groottes bó die vermoëns van sagtewarenutsgoed wat gewoonlik gebruik word om data binne ’n draaglike tydsduur vas te vang, saam te stel, te bestuur en te prosesseer.	Ngokuyintloko ibhekisa kwiiseti zedatha ezinkulu kakhulu okanye ezintsokothileyo ukuba zingasetyenzwa ngokwesiko lenkqubo yedatha yesoftware. Idatha enkulu idla ngokubandakanya iiseti zedatha ezinobungakanani obungaphaya kwamandla ezixhobo zesoftware ezisetyenziswa ngokuqhelekileyo ukubamba, ukulungelelanisa, ukulawula, nokucwangcisa idatha kwixesha elinyamezelekayo.	Kusho uhlu lwemininingwane olukhulu noma oludidayo oluzobhekana nesoftiwe yohlelo lokulungisa imininingwane lakudala. Imininingwane emikhulu ibandakanya uhlu lwemininingwane enobungakho obungaphezul kwamandla eamathuluzi esoftiwe evame ukusetshenziswa okuqopha, ukuphatha, ukulawula, nokulungisa imininingwane esikhathini esidlulile esamukelekile.
google_research_ai_terminology_53	Binary classification	binêre klassifisering	Ukuhlelwa kweBinary	ukuhlela kokukabili	A type of classification task that predicts one of two mutually exclusive classes: the positive class and the negative class. For example, a model that determines whether email messages are spam (the positive class) or not spam (the negative class).	’n Tipe klassifiseringtaak wat een of twee onverenigbare klassifiserings voorspel: die positiewe klas en die negatiewe klas. Byvoorbeeld, ’n model wat bepaal of ’n e-posboodskap gemorspos is (die positiewe klas) of nie gemorspos is nie (die negatiewe klas).	Uhlobo lomsebenzi wokuhlelwa oqikelela enye yeeklasi ezimbini ezizimeleyo: udidi oluhle kunye nodidi olubi. Umzekelo, imodeli emisela ukuba imiyalezo yeimeyile isispem (iklasi elungileyo) okanye ayisiso (udidi olubi).	Uhlobo lomsebenzi wokuhlukanisa ngokwamazinga oluqagela izinga elilodwa noma amabili akhethekile: amazinga okuhle kanye namazinga okubi. Isibonelo, uhlelo olunquma ukuthi ingabe imiyalelo yama-imeyli iwugaxekile (izinga lokuhle) noma akusiwo ugaxekile (izinga lokubi).
google_research_ai_terminology_54	Black Box AI	swartboks-KI	Ibhokisi yeAI eMnyama	Ubuhlakani Obakhiwe obungeqondwe abantu	A term used to describe artificial intelligence systems whose internal workings and decision-making processes are opaque or not easily understandable to humans.	’n Term wat gebruik word om KI-stelsels te beskryf waarvan die interne bewerkinge en besluitnemingprosesse ondeursigtig is en nie maklik deur mense verstaan kan word nie.	Ibinzana elisetyenziselwa ukuchaza iinkqubo zobukrelekrele obenziweyo apho iindlela zangaphakathi zokusebenza kwazo neenkqubo zazo zokwenza izigqibo zingacacanga okanye zingaqondwa lula ngabantu.	Itemu elisetshenziselwa ukuchaza ubuchwepheshe obusebenza ngokungasikho okwemvelo ezinezindlela zokusebenza zangaphakathi kanye nokuthatha izinqumo ezingacacile noma ezingaqondwa abantu kabula.
google_research_ai_terminology_55	Data minimization	dataminimalisering	unciphisozinkcukacha	Ukunciphisa imininingo	Data minimization is the principle of collecting, processing and storing only the necessary amount of personal information required for a specific purpose.	Dataminimalisering is die prinsiep waarin net die persoonlike inligting wat vir ’n spesifieke doel nodig is, ingesamel, verwerk en geberg word.	Ukunciphisa idatha ngumgaqo wokuqokelela, ukuprosesa nokugcina kuphela isixa esifunekayo senkcazelo yomntu efunekayo ngenjongo ethile.	Ukunciphisa imininingwane kuyinqubo yokuqoqa, ukulungisa nokugcina isibalo esidingekayo solwazi lomuntu siqu oludingeka ngokwenhloso ebaluliwe.
google_research_ai_terminology_56	Debiasing	vooroordeelverwydering	Ukunciphisa ukuthatha icala	Ukuchezuka kokujwayelekile	Debiasing is the reduction of bias, particularly with respect to judgment and decision making. Biased judgment and decision making is that which systematically deviates from the prescriptions of objective standards such as facts, logic, and rational behavior or prescriptive norms.	Vooroordeelverwydering is die vermindering van bevooroordeling, spesifiek met betrekking tot opinies en besluitneming. Bevooroordeelde opinies en besluitneming is dít wat stelselmatig afwyk van die voorgeskrewe objektiewe standaarde, soos feite, logikia en rasionele optrede of voorgeskrewe norme.	Ukunciphisa ukuthatha icala kukunciphisa umkhethe, ngakumbi ngokubhekisele kwisigwebo nokwenza izigqibo. Isigwebo nokwenziwa kwezigqibo okunomkhethe kukusuka ngokucwangcisiweyo kwimimiselo yemigangatho yenjongo efana neenyani, ingqiqo, nokuziphatha okunengqiqo okanye imigaqo emiselweyo.	Ukususa ukuchema kuwukunciphsa ukuchema, ikakhulukazi maqondana nokwahlulela kanye nokuthatha izinqumo. Ukwahlulela ngokungachemi kanye nokuthatha izinqumo kuyilokho okuchezuka ngokohlelo oluthile kulokho okumisiwe emigomweni yezimpokophelo njengamaqiniso, okuqondakalayo, kanye nokuziphatha ngokufanele noma izindlela ezimisiwe.
google_research_ai_terminology_57	Differential Privacy	differensiële privaatheid	Ukugcinwa kuyimfihlo kokwahlukileyo	Ubumfihlo Obuvikelayo	Differential privacy (DP) is a mathematically rigorous framework for releasing statistical information about datasets while protecting the privacy of individual data subjects. It enables a data holder to share aggregate patterns of the group while limiting information that is leaked about specific individuals.	Differensiële privaatheid (DP) is ’n wiskundige, strawwe raamwerk vir die vrystelling van statistiese inligting oor datastelle, terwyl die privaatheid van individuele data-onderwerpe beskerm word. Dit maak dat ’n datahouer aggregaatpatrone van die groep kan deel en terselfde tyd die inligting kan beperk wat deursyfer oor die spesifieke individue.	Ukugcinwa kokwahlukileyo kuyimfihlo (iDP) sisikhokelo semathematika esingqongqo sokukhupha inkcazelo yeenkcukacha manani ngeeseti zedatha ngelixa kukhusela ubumfihlo bedatha yento nganye. Kuvumela umnini wedatha ukuba abelane ngeepateni ezidityanisiweyo zeqela ngelixa enciphisa inkcazelo evuzisiweyo ngabantu abathile.	Inqubomogomo yokwahlukanisa (i-DP) iwuhlelo olunzulu lwezibalo lokukhipha ulwazi lwezibalo ngohlu lwemininingwane kube kuvikelwa imininingwane eyimfihlo yabantu abathile. Yenza umnikazi wemininingwane akwazi ukwabelana ngamaphethini ngokwezibalo ebe enciphisa ulwazi oluputshukayo ngabantu abathile.
google_research_ai_terminology_58	Edge Computing	periferale berekening	Usondezomphethelweni ngekhompuyutha	Ucubungulokusondeza	Edge computing is a distributed computing model that brings computation and data storage closer to the sources of data. Edge is about processing data closer to where it's being generated, enabling processing at greater speeds and volumes, leading to greater action-led results in real time.	Periferale berekening is ’n verspreide berekeningmodel wat berekeninge en databerging nader aan databronne bring. Periferaal behels dat data geprosesseer word nader aan waar dit gegenereer word. Dit maak dit moontlik om teen hoër spoed en groter volumes data te prosesseer, en daardeur intyds beter handelinggedrewe resultate te kry.	Ukusetyenziswa kwekhompyutha emphethelweni yimodeli yekhompyutha esasazwayo ezisa ukusetyenziswa kwekhompyutha nokugcinwa kwedatha kufutshane nemithombo yedatha. Emphethelweni umalunga nokusetyenzwa kwedatha kufutshane nalapho yenziwa khona, ivumela ukuproseswa ngezantya ezinkulu nemithamo, ekhokelela kwiziphumo ezinkulu ezikhokelwa yintshukumo ngexesha lokwenyani.	Ucubungulokuzondeza lwukudlulisela uhlelo locwaningo lwamakhompyutha olusondeza ucwaningo nokugcinwa kolwazi ndawonye nemithombo yemininingwane. Ucubungulokusondeza lumayelana nokulungisa imininingwane ibe iseduze kwalapho ikhiqizwe khona, okwenza ukulungiswa kwayo kusheshe futhi ibe miningi, okuholela emiphumeleni ebanzi ngesikhathi sangempela.
google_research_ai_terminology_59	False negative	vals negatief	Inegethivu engeyonyani	Iphutha okungelona	A result that indicates a given condition does not exist when it does. This means the algorithm generates a negative result when it should have been positive. A false negative could result in a cancerous region being overlooked, leading to a delayed diagnosis and treatment.	’n Uitslag wat aandui dat ’n gegewe kondisie nie bestaan nie terwyl dit wel bestaan. Dit beteken dat die algoritme ’n negatiewe resultaat genereer wanneer dit ’n positiewe een moes genereer. Byvoorbeeld, ’n vals negatief kan veroorsaak dat ’n kankeragtige area nie raakgesien word nie, wat kan lei tot vertraagde diagnose en behandeling.	Isiphumo esibonisa ukuba imeko echaziweyo ayikho kodwa ibe ikhona. Oku kuthetha ukuba ialgorithm ivelisa iziphumo ezinegethivu xa bekufanele ukuba ziphozithivu. Inegethivu engoyonyani inokukhokelela ekubeni ummandla onomhlaza ungahoywa, nto leyo ekhokelela kuxilongo nonyango olulibazisekileyo.	Umphumela oveza ukuthi isimo esithile asikho sibe sikhona. Lokhu kusho ukuthi i-algorithm ikhiqiza umphumela omubi kube kumele ikhiphe omuhle. Umphumelakukhohlisa ungadala ukuthi indawo enomdlavuza inganakwa, okuholela ekuphuzeni kokutholakala nokwelashwa kwesifo.
google_research_ai_terminology_60	False positive	vals positief	Iphozithivu engeyonyani	Iphutha okuyilo	A result that indicates a given condition does exist when it does not. For instance, in a medical imaging scenario, a false positive might lead to a healthy region being flagged as cancerous, potentially causing unnecessary concern and medical procedures for the patient.	’n Uitslag wat aandui dat ’n gegewe kondisie bestaan terwyl dit nie bestaan nie. Byvoorbeeld, in ’n mediesebeeldscenario kan ’n vals positief veroorsaak dat ’n gesonde area gevlag word as kankeragtig, wat onnodige bekommernis en mediese prosedures vir die pasient kan veroorsaak.	Isiphumo esibonisa ukuba imeko echaziweyo ikhona kodwa ibe ingekho. Umzekelo, kwimeko yomfanekiso wezonyango, iphozithivu ongeyonyani unokukhokelela ekubeni ummandla osempilweni kuthiwe unomhlaza, nto leyo enokubangela inkxalabo neenkqubo zonyango ezingeyomfuneko kwisigulana.	Umphumela oveza ukuthi isimo esibaluliwe sikhona sibe singekho. Isibonelo, lapho kuskenwa umzimba ukuze welashwe, umphumela wokuhle oyinkohliso ungaholela ekutheni indawo ephilile kuthiwe inomdlavuza, okungadala ukukhathazeka kanye nequbo yokwelashwa kwesidingo kungenasidingo.
google_research_ai_terminology_61	Federated Learning	gefedereerde leer	Ukufunda okudityanisiweyo	Ukuthuthukisa ubuchwepheshe	A distributed machine learning approach that trains machine learning models using decentralized examples residing on devices such as smartphones. In federated learning, a subset of devices downloads the current model from a central coordinating server. The devices use the examples stored on the devices to make improvements to the model. The devices then upload the model improvements (but not the training examples) to the coordinating server, where they are aggregated with other updates to yield an improved global model.	’n Verspreide masjienleerbenadering wat masjienleermodelle oplei met gedesentraliseerde voorbeelde wat op toestelle soos slimfone gestoor is. In gefedereerde leer laai ’n substel toestelle die huidige model van ’n sentrale, koördineringbediener af. Die toestelle gebruik die voorbeelde wat op die toestelle gestoor is om verbeteringe aan die model te maak. Die toestelle laai dan die modelverbeteringe op (maar nie die opleidingvoorbeelde nie) na die koördineringbediener toe, waar dit alles saam met ander opdaterings saamgevoeg word om ’n verbeterde algehele model te lewer.	Indlela yokufunda koomatshini esasazwayo eqeqesha iimodeli zokufunda ngoomatshini kusetyenziswa imizekelo ebekiweyo ehlala kwizixhobo ezinjengeesmartphone. Kwimfundo edibeneyo, iseti engezantsi yezixhobo ikhuphela imodeli yangoku kwiseva ephakathi yokulungelelanisa. Izixhobo zisebenzisa imizekelo egcinwe kwizixhobo ukuze kuphuculwe imodeli. Emva koko izixhobo zifaka ukuphuculwa kwemodeli (kodwa kungekhona imizekelo yoqeqesho) kwiseva yokulungelelanisa, apho zidityaniswe nolunye uhlaziyo ukuvelisa imodeli ephuculweyo yehlabathi.	Indlela yokufunda ngomshini edluliswayo eqeqesha izinhlelo zokufunda ngemishini kusetshenziswa izibonelo eziqeqelwe ndawonye kumadivayisi anjengamasamthifoni. Lapho kufundwa ngokubuchwepheshe, uhlu lwamadivayisi akhethiwe adawuniloda uhlelo olukhona ngokususela kuseva edidiyele eqoqele konke ndawonye. Amadivayisi asebenzisa izibonelo ezigcinwe kumadivayisi ukuthuthukisa uhlelo. Amadivayisi abe esefaka okuthuthukisiwe ohlelweni (kodwa hhayi isibonelo zokuqeqesha) kuseva edidiyelayo, lapho ikalwe khona nokunye ukuthuthukisa ukuze kuphume uhlelo oluthuthukisiwe lomhlaba jikelele.
google_research_ai_terminology_62	Feedback Loop	terugvoerkring	Iluphu yengxelo	Iwuku Lezimvo	Feedback occurs when outputs of a system are routed back as inputs as part of a chain of cause-and-effect that forms a circuit or loop. The system can then be said to feed back into itself. For example, if the AI system is a chatbot, it might receive feedback from users on how well it is able to understand and respond to their queries. The AI system then uses this feedback to adjust its algorithms and improve its performance.	Terugvoering vind plaas wanneer ’n stelsel se uitsette terugherlei word as insette as deel van ’n oorsaak-en-gevolgkettingverloop wat dan ’n kring of lusloop veroorsaak. Daar kan dus gesê word dat die stelsel in ditself in terug voer. Byvoorbeeld, as die KI-stelsel ’n kletsbot is, kan dit dalk terugvoer kry vannaf gebruikers oor hoe goed dit hul navrae kan verstaan en daarop reageer. Die KI-stelsel gebruik dan hierdie terugvoer om sy algoritmes aan te pas en sy werkverrigting te verbeter.	Ingxelo yenzeka xa iziphumo zenkqubo zibuyiselwa umva njengenxalenye yekhonkco lesizathu nesiphumo esenza isekethe okanye iluphu. Ngoko kunokuthiwa inkqubo ibuyela kuyo ngokwayo. Umzekelo, ukuba inkqubo ye-AI yi-chatbot, inokufumana impendulo kubasebenzisi malunga nokuba ikwazi kangakanani ukuqonda nokuphendula imibuzo yabo. Emva koko, inkqubo yeAI isebenzisa le ngxelo ukulungisa iialgorithm zayo nokuphucula ukusebenza kwayo.	Ukuthola izimvo kwenzeka lapho okufakiwe ohlelweni kubuyiselwa njengezimvo njengengxenye yochungechunge lokwenzekile nomphumela olwakha isekhethi noma iwuku. Uhlelo lungathathwa ngokuthu luyizimvo ngokwalo. Isibonelo, uma uhlelo lwe-AI luyichatbot, lungase luthola izimvo ezivela kubasebenzisi ngokuthi lukwazi kangakanani ukuqonda nokuphendula imibuzo yalo. Uhlelo lwe-AI lube selusebenzisa lezi zimvo ukuchibiyela ama-algorithm futhi kuthuthukiswe ukusebenza kwawo.
google_research_ai_terminology_63	Hyperparameter	hiperparameter	Ihyperparameter	Ilungiselelakuqeqesha	The variables that you or a hyperparameter tuning service adjust during successive runs of training a model. For example, learning rate is a hyperparameter. You could set the learning rate to 0.01 before one training session. If you determine that 0.01 is too high, you could perhaps set the learning rate to 0.003 for the next training session.	Die veranderlikes wat jy of ’n hipergrensinstellingdiens aanpas tydens agtereenvolgende opleidingsessies van ’n model. Byvoorbeeld, leerkoers is ’n hipergrens. Jy kan die leerkoers op 0.01 stel voor een opleidingsessie. As jy bepaal dat 0.01 te hoog is, kan jy die leerkoers dan dalk na 0.003 toe stel vir die volgende opleidingsessie.	Izinto eziguquguqukayo ozihlengahlengisa ngokwakho okanye yinkonzo yehyperparameter tuning ngexesha loqeqesho olulandelelanayo lwemodeli. Umzekelo, izinga lokufunda yi-hyperparameter. Unokuseta ireyithi yokufunda ibe yi-0.01 phambi kweseshoni yoqeqesho enye. Ukuba ufumanisa ukuba u-0.01 uphezulu kakhulu, usenokuseta ireyithi yokufunda ibe ngu-0.003 kwiseshoni yoqeqesho elandelayo.	Izakhi ezisetshenziswa nguwe noma ihyperparameter ngasikhathi lapho uqeqesha uhlelo nempumelelo, Isibonelo, izinga lokufunda liyihyperparameter. Ungahlela izinga lokufunda libe u-0.01 ngaphambi kwesikhathi sokufundisa esisodwa. Uma unquma ukuthi u-0.01 uphezulu kakhulu, ungase umise izinga lokufunda ku-0.003 esikhathini sokuqeqesha esilandelayo.
google_research_ai_terminology_64	Image recognition	beeldherkenning	Ukuqondwa komfanekiso	Ukubona isithombe	Also known as image classification, is a process that classifies object(s), pattern(s), or concept(s) in an image.	Ook bekend as prentklassifisering. Dit is ’n proses wat voorwerp(e), patroon(e), of konsep(te) in ’n prent klassifiseer.	Ikwaziwa njengohlelo lomfanekiso, yinkqubo ehlela into (izinto), ipatheni (iipatheni), okanye inkalo (iinkalo) kumfanekiso.	Kubuye kwaziwe ngokuthi uhlelo lokubeka izithombe ngokwamazinga, inqubo ebeka izinto, amaphethini, noma imiqondo ngokwamazinga esithonjeni.
google_research_ai_terminology_65	Intelligent agent	intelligente agent	Iarhente yobukrelekrele	Isimeleli esikhaliphile	IA / An agent that perceives its environment, takes actions autonomously in order to achieve goals, and may improve its performance with learning or acquiring knowledge. An intelligent agent may be simple or complex e.g. a thermostat, a human being...	IA / ’n Agent wat sy omgewing waarneem, voer selfstandig handelinge uit om doelwitte te bereik, en kan sy werkverrigting verbeter deur te leer of kennis in te win. ’n Intelligente agent kan eenvoudig of kompleks wees, byvoorbeeld ’n termostaat, ’n mens …	IA / Iarhente ebona okuyingqongileyo, ithatha amanyathelo ngokuzimeleyo ukuze ifezekise iinjongo, kwaye inokuphucula ukusebenza kwayo ngokufunda okanye ngokufumana ulwazi. Iarhente ekrelekrele isenokuba lula okanye intsonkothe umz. ithermostat, umntu..	Isimeleli esiqonda indawo esikuyo, esithatha isinyathlo ngokuzenzekela ukuze kuveze imogomo, futhi kuthuthukiswe ukusebenza ngokufunda noma ukuthola ulwazi. Isimeleli esikhaliphile singaba lula noma nzulu isib. isiklalizingakushisa, umuntu
google_research_ai_terminology_66	Memory leak	geheuelek	Ukuvuza kwememori	Ukuputshuka kolwazi olugciniwe	In computer science, a memory leak is a type of resource leak that occurs when a computer program incorrectly manages memory allocations in a way that memory which is no longer needed is not released. It reduces the performance of the computer by reducing the amount of available memory.	In rekenaarwetenskap is geheuesyfering ’n tipe hulpbronverlies wat voorkom wanner ’n rekenaarprogram verkeerdelik geheuetoewysing bestuur op só ’n wyse dat geheue wat nie meer benodig word nie, nie vrygestel word nie. Dit verminder die werkverrigting van die rekenaar deur die beskikbare geheue te verminder.	Kwisayensi yekhompyutha, ukuvuza kwememori luhlobo lokuvuza kwezixhobo okwenzekayo xa inkqubo yekhompyutha ilawula ngokungalunganga ukwabiwa kweememori ngendlela yokuba imemori engasadingekiyo ingakhutshwa. Kunciphisa ukusebenza kwekhompyutha ngokunciphisa inani lememori ekhoyo.	Kwezesayensi yamakhompyutha, ukuputshuka kolwazi kuwuhlobo lokuputshuka kwemithombo okwenzeka lapho uhlelo lwekhompyutha lulawula ukwabiwa kolwazi olugciniwe ngokungafanele ngendlela yokuthi ulwazi olugciniwe olungasadingeki lungakhishwa. Lunciphisa ukusebenza kwekhompyutha ngokunciphisa ubungako bolwazi olugciniwe olukhona.
google_research_ai_terminology_67	Overfitting	oormatige passing	Ukungalingani okugqithileyo	Ukuhambisana ngqo	In mathematical modeling, overfitting is "the production of an analysis that corresponds too closely or exactly to a particular set of data, and may therefore fail to fit to additional data or predict future observations reliably"	In wiskundige modelleerwerk is oorpassing “die produsering van ’n ontleding wat te nou of presies ooreenstem met ’n datastel en daarom nie betroubaar by bykomende data kan pas of toekomstige waarnemens kan voorspel nie.	Kwimodeli yemathematika, ukulingana ngokugqithisileyo "kukuveliswa kohlalutyo oluhambelana ngokusondeleyo okanye ngokuthe ngqo kwiseti ethile yedatha, kwaye ke ngoko inokusilela ukulingana nedatha eyongezelelweyo okanye ukuqikelela imigqaliselo yexesha elizayo ngokuthembekileyo"	Ohlelweni lwezibalo, ukuhambisana ngqo "kuwukukhiqizwa kocwaningo oluhambisana kakhulu noma ngqo nohlu lweminingwane ethile, futhi olungase lwahluleke ukungena kweminye imininingwane noma ukuqagela okuzohlolwa ngokuzayo ngokwethembeka"
google_research_ai_terminology_68	Oversampling	opbalansering	Iisampuli egqithisileyo	Ukusetshenziswa kwamasampula kabusha	Reusing the examples of a minority class in a class-imbalanced dataset in order to create a more balanced training set.	Die hergebruik van voorbeelde van ’n minderheidklas in ’n klasongebalanseerde datastel om sodoende ’n meer gebalanseerde opleidingstel te skep.	Ukusebenzisa kwakhona imizekelo yeklasi encinci kwiseti yedatha kwiklasi yokungalingani ukwenzela ukudala isethi yoqeqesho olulungeleleneyo.	Ukusebenzisa kabusha izibonelo ezingeni lokuncane ohlwini lwemininingwane olungalingene ngokufanele ukuze kwakhiwe uhlu lokuqeqesha olulingene.
google_research_ai_terminology_69	Precision	presisie	Ukuchana	ukuqonda ngqo	Precision is the fraction of relevant instances among the retrieved instances. It is calculated by dividing relevant retrieved instances with all retrieved instances. For example, a computer program for recognizing dogs (the relevant element) in a digital photograph processes a picture which contains 10 cats and 12 dogs. The program then identifies 8 dogs. Of the eight elements identified as dogs, only 5 actually are dogs (true positives), while the other 3 are cats (false positives). 7 dogs were missed (false negatives), and 7 cats were correctly excluded (true negatives). The program's precision is then 5/8 (true positives / selected elements).	Presisie is die breukdeel van relevante gevalle uit die gerapporteerde gevalle. Dit word uitgewerk deur die relevante gerapporteerde gevalle te deel deur al die gerapporteerde gevalle. Byvoorbeeld, ’n rekenaarprogram wat honde (die relevante element) in digitale foto's identifiseer, prosesseer ’n foto wat 10 katte en 12 honde in het. Die program identifiseer dan 8 honde. Uit die 8 elemente wat as honde geïdentifiseer is, is net 5 regtig honde (werklike positief) , terwyl 3 katte is (vals positief). 7 honde is misgekyk (vals negatief) en 7 katte is korrek uitgesluit (werklike negatief). Die program se presisie is dan 5/8 (werklike positiewe / gerapporteerde elemente).	Ukuchaneka liqhezu leemeko ezifanelekileyo phakathi kweemeko ezifunyenweyo. Ibalwa ngokwahlula iimeko ezifanelekileyo kwezifunyenweyo nazo zonke iimeko ezifunyenweyo. Umzekelo, inkqubo yekhompyutha yokuqaphela izinja (into efanelekileyo) kwifoto yedijithali iprosesa umfanekiso oqulethe iikati ezili-10 nezinja eziyi-12. Inkqubo ke iye ichonge izinja ezisi-8. Kwizinto ezisibhozo ezichongwe njengezinja, zi-5 kuphela ezizinja (iiphozithivu zokwenyani), ngelixa ezinye ezi-3 izikati (iiphozithivu ezingeyonyani). Izinja ezisi-7 ziphosiwe (iinegethivu ezingeyonyani), zaze iikati ezisi-7 zangabandakanywa ngokuchanekileyo (iinegethivu eziyinyani). Ukuchaneka kwenkqubo ngu-5/8 (iiphozithivu eziyinyani / izinto ezikhethiweyo).	Ukugqonda ngqo kuyingcosana yezimo ezifanele phakathi kwezimo ezilandiwe. Kubalwa ngokuhlukanisa izimo ezilandiwe ezifanele nazo zonke izimo ezilandiwe. Isibonelo, uhlelo lwekhompyutha lokubona izinja (isakhi esifanele) esithombeni sedijithali izinqubo zesithombe eziqukethe amakati ayi-10 nezinja eziyi-12. Uhlelo lube seluhlonza izinja eziyi-8. Ezakhini eziyisishiyagalombili ezihlonzwe njengezinja, zi-5 kuphela eziyizinja (umphumela wokuhle oyiqiniso), kube ezi-3 zingamakati (umphumela wokhlei oyinkohliso) Izinja eziyisi-7 azibonwanga (imiphumela yokubi eyinkohliso), kanye namakati ayisi-7abandakanywe ngokufanele (umphumela wokubi oyiqiniso) Ukuqonda ngqo kohlelo lube seluba ngu-5/8 (imiphumela emihle ngokweqiniso / izakhi ezikhethiwe).
google_research_ai_terminology_70	Predictive analytics	voorspelontleding	Uhlalutyo oluqikelelayo	Ucubungulo oluqagelayo	Predictive analytics is a form of analytics that uses machine learning or technology to predict what will happen in a specific time frame based on historical data and patterns.	Voorspelontleding is ’n vorm van ontleding wat masjienleer of tegnologie gebruik om volgens historiese data en patrone te voorspel wat in ’n spesifieke tydperk sal gebeur.	Uhlalutyo lokuqikelela luhlobo lohlalutyo olusebenzisa ukufundwa koomatshini okanye iteknoloji ukuze kuqikelelwe oko kuza kwenzeka kwixesha elithile ngokusekelwe kwidatha yembali neepatheni.	Ucubungulo oluqagelayo wuhlobo lwezibalo olusebenzisa ukufunda ngemishini noma ubuchwepheshe ukuqagela okuzokwenzeka esikhathi esithile esimisiwe kususelwa kumininingwane esemlandweni kanye namaphethini.
google_research_ai_terminology_71	Privacy Impact Assessment	privaatheidimpakassessering	UVavanyo lweMpembelelo yoBumfihlo	Ukuhlola Umthelela Wobumfihlo	PIA / A Privacy Impact Assessment (PIA) is a process which assists organizations in identifying and managing the privacy risks arising from new projects, initiatives, systems, processes, strategies, policies, business relationships etc. / <A Privacy Impact Assessment is a process which assists organizations in identifying and managing the privacy risks arising from new projects, initiatives, systems, processes, strategies, policies, business relationships etc.>	PIA / ’n Privaatheidimpakassessering (PIA) is ’n proses wat organisasies help om die privaatheidrisiko's te identifiseer en te bestuur wat kan voortspruit uit nuwe projekte, innisiatiewe, stelsels, prosesse, strategieë, beleide, besigheidsverhoudings ens. / <’n Privaatheidimpakassessering is ’n proses wat organisasies help om die privaatheidrisiko's te identifiseer en te bestuur wat kan voortspruit uit nuwe projekte, innisiatiewe, stelsels, prosesse, strategieë, beleide, besigheidsverhoudings ens.>	iPIA / Uvavanyo Lwempembelelo Yobumfihlo (PIA) yinkqubo enceda imibutho ekuchongeni nasekulawuleni imingcipheko yobumfihlo evela kwiiprojekthi ezintsha, izinto ezintsha, iisistim, iinkqubo, izicwangciso, iipolisi, ubudlelwane boshishino njl.njl.	Ukuhlola Umthelela Wobumfihlo (i-PAIA) kuyinqubo esiza izinhlangano ekuhlonzeni nasekuphatheni ingcuphe yobumfihlo evela kumaphrojekthi amasha, izinhlelo kanye nezinqubo, amaqhingasu, izinqubomgomo, ubudlelwano bebhizinisi njl.
google_research_ai_terminology_72	Prompt injection	instruksieteksinvoeging	Utofomyalelo	Ukufaka imiyalelo	An attack that manipulates a large language model (LLM) by injecting malicious inputs designed to alter the model’s output. This type of cyber-attack exploits the way LLMs process and generate text based on input prompts. By inserting carefully crafted text, an attacker can trick the model into producing unauthorized content, accessing restricted data, or executing specific actions.	’n Aanval wat ’n groottaalmodel (GTM) manipuleer deur kwaadwillige invoere in te voeg wat ontwerp is om die model se uitset te verander. Hierdie soort kuberaanval byt die manier uit waarop GTM'e teks met invoerporboodskappe prosesseer en genereer. Deur fyn saamgestelde teks in te voeg, kan ’n aanvaller die model flous om dit ongoedgekeurde inhoud te laat produseer, toegang te kry tot beperkte data, of om spesifieke handelinge uit te voer.	Uhlaselo olusebenzisa imodeli yolwimi olukhulu (LLM) ngokutofa igalelo elibi eliyilelwe ukuguqula imveliso yemodeli. Olu hlobo lohlaselo lwecyber lusebenzisa indlela iiLLM eziprosesa zize zivelise ngayo itekisi ngosekelwe kumyalelo ofakiweyo. Ngokufaka itekisi eyenziwe ngobuchule, umhlaseli unokukhohlisa imodeli ukuba ivelise ikhontenti engagunyaziswanga, ifikelele kwidatha ethintelweyo, okanye ukwenza izenzo ezithile.	Ukuhlasela okusebenzisa izinhlelo zolimi ezinkulu (ama-LLM) ngokufaka okumoshayo okwakhele ukuphazamisa umphumela wohlelo. Lolu hlobo lokuhlasela nge-inthanethi kuxhaphaza indlela ama-LLM alungisa futhi akhiqize ngayo izibizelo zokufakwayo. Ngokufaka ngokucophelela umbhalo obunjiwe, umhlaseli angadida uhlelo ukuze lukhiqize ulwazi olungagunyazwanga, lufinyelele emininingwaneni evinjelwe, noma lusebenise imiyalelo emisiwe.
google_research_ai_terminology_73	Pseudonymization	pseudonimisering	IPseudonymization	Ukuzivezambumbulu	The process of obscuring data with the ability to re-identify it later. Personally identifiable information fields within a data record are replaced by one or more artificial identifiers, or pseudonyms.	Die proses om data te verdoesel met die vermoë om dit later te heridentifiseer. Velde met persoonlik identifiseerbare inligting in ’n datarekord word vervang met een of meer kunsmatige identifiseerders, of pseudonieme.	Inkqubo yokufihla idatha nokwazi ukuyichonga kwakhona kamva. Indawo yenkcazelo echongiweyo ngokobuqu kwirekhodi yedatha ethathelwe indawo sisifanisi esinye, okanye ngapezulu, okanye ii-pseudonym.	Inqubo yokufihla idatha ngendlela yokukwazi ukuyibona futhi kamuva. Izinkundla zolwazi olukhomba umuntu siqu (PII) erekhodini ledatha ziyakhishwa kufakwe izinkomba zokwenziwa ezizodwa noma ngaphezulu, noma amapseudonyms.
google_research_ai_terminology_74	Recall	volledigheid	Ukukhumbula	Isibuyiseli	Recall is the fraction of relevant instances that were retrieved. It is calculated by dividing relevant retrieved instances with all relevant instances. For example, a computer program for recognizing dogs (the relevant element) in a digital photograph processes a picture which contains 10 cats and 12 dogs. The program then identifies 8 dogs. Of the eight elements identified as dogs, only 5 actually are dogs (true positives), while the other 3 are cats (false positives). 7 dogs were missed (false negatives), and 7 cats were correctly excluded (true negatives). The program's recall is 5/12 (true positives / relevant elements).	Volledigheid is die breukdeel van relevante gevalle wat gerapporteer is. Dit word uitgewerk deur die relevante gerapporteerde gevalle te deel deur die totale aantal relevante gevalle. Byvoorbeeld, ’n rekenaarprogram wat honde (die relevante element) in digitale foto's identifiseer, prosesseer ’n foto wat 10 katte en 12 honde in het. Die program identifiseer dan 8 honde. Uit die 8 elemente wat as honde geïdentifiseer is, is net 5 regtig honde (werklike positief) , terwyl 3 katte is (vals positief). 7 honde is misgekyk (vals negatief) en 7 katte is korrek uitgesluit (werklike negatief). Die program se volledigheid is dan 5/12 (werklike positiewe / relevante elemente).	Ukukhumbula liqhezu leemeko ezifanelekileyo eziye zafunyanwa. Kubalwa ngokwahlula iimeko ezifanelekileyo kwezifunyenweyo nazo zonke iimeko ezifanelekileyo. Umzekelo, inkqubo yekhompyutha yokuqaphela izinja (into efanelekileyo) kwifoto yedijithali iprosesa umfanekiso oqulethe iikati ezili-10 nezinja eziyi-12. Inkqubo ke iye ichonge izinja ezisi-8. Kwizinto ezisibhozo ezichongwe njengezinja, zi-5 kuphela ezizinja (iiphozithivu zokwenyani), ngelixa ezinye ezi-3 izikati (iiphozithivu ezingeyonyani). Izinja ezisi-7 ziphosiwe (iinegethivu ezingeyonyani), zaze iikati ezisi-7 zangabandakanywa ngokuchanekileyo (iinegethivu eziyinyani). Ukukhumbula kwenkqubo ngu-5/12 (iiphozithivu eziyinyani / izinto ezifanelekileyo).	Ukubuyisa kuyingcosana yezimo ezifanele ezilandiwe. Kubalwa ngokuhlukanisa izimo ezilandiwe ezifanele nazo zonke izimo ezifanele. Isibonelo, uhlelo lwekhompyutha lokubona izinja (isakhi esifanele) esithombeni sedijithali izinqubo zesithombe eziqukethe amakati ayi-10 nezinja eziyi-12. Uhlelo lube seluhlonza izinja eziyi-8. Ezakhini eziyisishiyagalombili ezihlonzwe njengezinja, zi-5 kuphela eziyizinja (umphumela wokuhle oyiqiniso), kube ezi-3 zingamakati (umphumela wokhlei oyinkohliso) Izinja eziyisi-7 azibonwanga (imiphumela yokubi eyinkohliso), kanye namakati ayisi-7abandakanywe ngokufanele (umphumela wokubi oyiqiniso) Ukukhumbula kohlelo kube ngu-5/12 (imiphumela emihle ngokweqiniso / izakhi ezifanele).
google_research_ai_terminology_75	Reinforcement Learning from human feedback	versterkingleer uit menslike terugvoer	Ukufunda okukhuthazwa yingxelo yabantu	Ukufunda Ukuthatha Izinqumo ngebuyambiko elivela kubantu	In machine learning, reinforcement learning from human feedback (RLHF) is a technique to align an intelligent agent with human preferences. It involves training a reward model to represent preferences, which can then be used to train other models through reinforcement learning.	In masjienleer is verstekleer uit menseterugvoer (VLMT) ’n tegniek om ’n intelligente agent te belyn met mensevoorkeure.	Ekufundweni koomatshini, ukufunda okukhuthazwa yingxelo esuka kubantu (iRLHF) bubuchule bokulungelelanisa iarhente ekrelekrele nezinto ezikhethwa ngabantu. Ibandakanya ukuqeqesha imodeli yokuvuza ukuze imele izinto ezikhethwayo, ezinokuthi zisetyenziselwe ukuqeqesha ezinye iimodeli ngokufunda okukhuthazayo.	Ekufundeni kwemishini, ukufunda kokugcizelela okususelwa ezimvweni zabantu (i-RLHF) kuyindlela yokuqondanisa isimeleli esikhaliphile nalokho okukhethwa abantu. Kubandakanya ukuqeqesha uhlelo lwemiphumela ukuze luveze okukhethwayo, okungasetshenziselwa ukuqeqesha ezinye izindlela zokugcizelela ukufunda.
google_research_ai_terminology_76	Retrieval augmented generation	ophaalaangevulde generering	Ukuvelisa ukufumana okwandisiweyo	ukukhiqiza kokulanda ulwazi	Retrieval augmented generation (RAG) is a type of generative artificial intelligence that has information retrieval capabilities. It modifies interactions with a large language model (LLM) so that the model responds to user queries with reference to a specified set of documents, using this information in preference to information drawn from its own vast, static training data. This allows LLMs to use domain-specific and/or updated information.	Aanvullendeophaalgenerering (AOG) is ’n tipe generatiewe kunsmatige intelligensie wat inligtingophaalvermoëns het. Dit verander interaksies met ’n groottaalmodel (GTM) sodat die model reageer op navrae met verwysing na ’n gespesifiseerde stel dokumente en hierdie inligting met voorkeur gebruik in plaas van inligting wat dit vanuit sy eie wye, onveranderde opleidingdata trek. Dit maak dat die GTM domeinspesifieke en/of opgedateerde inligting kan gebruik.	Ukuvelisa ukufumana okwandisiweyo (iRAG) luhlobo lobukrelekrele obenziweyo obuvelisayo obunesakhono sokufumana inkcazelo. Ilungisa unxibelelwano nemodeli yolwimi olukhulu (LLM) ukuze imodeli iphendule imibuzo yabasebenzisi ngokubhekisele kuluhlu oluthile lwamaxwebhu, isebenzisa le nkcazelo ngokukhetha inkcazelo ethathwe kwidatha yayo enkulu, engatshintshiyo yoqeqesho. Oku kuvumela iiLLM ukuba zisebenzise idomeyini ekhethekileyo kunye/okanye inkcazelo ehlaziyiweyo.	Ukukhiqiza kokulnadna ulwazi (i-RAG) kuwuhlobo lobuchwepheshe obusebenza ngokungasikho okwemvelo obukhiqizayo olukwazi ukulanda ulwazi. Lulungisa ukuxhumana kwezinhlelo zolimi ezinkulu (ama-LLM) ukuze izinhlelo ziphendule imibuzo yomsebenzisi zisusela ulwazi ohlwini lwemibhalo, zisebenzisa lolu lwazi kunokukhetha ulwazi oluthathwe emininingwaneni yalo yokuqeqesha ebanzi ngemile. Lokhu kwenza ukuthi ama-LLM akwazi ukusebnezisa ulwazi oluqondene nomkhakha othile kanye/noma olubuyekeziwe.
google_research_ai_terminology_77	Robustness	robuustheid	Ukomelela	Ubuqatha	In computer science, robustness is the ability of a computer system to cope with errors during execution and cope with erroneous input. It is tested using techniques such as fuzz testing (an automated software testing technique that involves providing invalid, unexpected, or random data as inputs to a computer program. The program is then monitored for exceptions such as crashes or potential memory leaks.)	In rekenaarwetenskap is robuustheid die vermoë van ’n rekenaarstelsel om foute tydens uitvoering en foutiewe invoer te kan hanteer. Dit word getoets met tegnieke soos donstoetse (’n outomatiese sagtewaretoetstegniek waarin ongeldige, onverwagte of lukrake data as invoere vir ’n rekenaarprogram verskaf word. Die program word dan gemonitor vir uitsonderings soos omvalle of moontlike geheuesyferings).	Kwisayensi yekhompyutha, ukomelela kukukwazi kwenkqubo yekhompyutha ukumelana neengxaki ngexesha lokwenziwa nokujongana negalelo elinengxaki. Ivavanywa kusetyenziswa ubuchule obufana novavanyo lokufana (ubuchule bokuvavanya isoftwe ezenzekelayo ebandakanya ukunika idatha engasebenziyo, engalindelekanga, okanye engacwangciswanga njengegalelo kwinkqubo yekhompyutha. Inkqubo iye ibekwe iliso kwizinto ezisecaleni ezifana nokukhresha okanye ukuvuza kwememori okunokubakho.)	Kwezesayensi yekhompyutha, ukubekeza kungamandla ohlelo lwekhompyutha lokubekezelela amaphutha lapho lusetshenziswa futhi lubekezelele okufakwe ngephutha. Luhlolwa kusetshenziswa izindlela ezinjengokuhlola ngokungaphelele (indlela yokuhlola ngesoftiwe ngokuzenzekela okubandakanya ukuhlinzeka ngemininingwane engasebenzi, engalindelekile, noma engavamile njengokufakwa ohlelweni lwekhompyutha. Uhlelo lube seluqashwa kubhekwe okuphuma eceleni njengokushayisa noma ukuputshuka kolwazi okungase kube khona.)
google_research_ai_terminology_78	Self-training	selfopleiding	Ukuziqeqeshela	Ukuziqeqesha	A variant of self-supervised learning that is particularly useful when all of the following conditions are true: - The ratio of unlabeled examples to labeled examples in the dataset is high. - This is a classification problem.	’n Variant van self-gekontrolleerde leer wat besonders nuttig is wanneer al die volgende toestannde waar is: - Die verhouding van ongeëtiketteerde voorbeelde teenoor geëtiketteerde voorbeelde in die datastel hoog is. - Dit ’n klassifiseringprobleem is.	Umahluko wokufunda okugadwe ngokuphakathi uluncedo ngakumbi xa zonke ezi meko zilandelayo ziyinyani: - Umlinganiselo wemizekelo engabhalwanga kwimizekelo enelebhile kuluhlu lwedatha uphezulu. - Le yingxaki yokuhlelwa.	Igatsha lokuzifundela oluwusiza lapho okulandelayo kuyiqiniso: - Lokhu kuyinkinga yokubeka ngokwamazinga.
google_research_ai_terminology_79	Superintelligence	superintelligensie	Ubukrelekrele obuphezulu	Ubuhlakani obukhulukazi	A superintelligence is a hypothetical agent that possesses intelligence surpassing that of the brightest and most gifted human minds.	’n Superintelligensie is ’n hipotetiese agent wat intelligensievermoëns het wat ver bo dié van die slimste en mees begaafde menslike denke is.	Ubukrelekrele obuphezulu yiarhente ecingelwayo enobukrelekrele obudlula ezona ngqondo ziqaqambileyo nezinesiphiwo zomntu.	Ubuhlakani obukhulukazi buyisimeleli esingasiso esangempela esiqukethe ubuhlakani obudlula abantu abahlakaniphe kakhulu kakhulu.
google_research_ai_terminology_80	Temperature	temperatuur	Ubushushu	Iqaphasimo	A hyperparameter that controls the degree of randomness of a model's output. Higher temperatures result in more random output, while lower temperatures result in less random output. Choosing the best temperature depends on the specific application and the preferred properties of the model's output. For example, you would probably raise the temperature when creating an application that generates creative output. Conversely, you would probably lower the temperature when building a model that classifies images or text in order to improve the model's accuracy and consistency.	’n Hipergrens wat die mate van lukraakheid van ’n model se uitset beheer. Hoër temperature veroorsaak meer lukrake uitsette, terwyl laer temperature minder lukrake uitsette veroorsaak. Om die beste temperatuur te kies hang af van die spesifieke toepassing en die voorkeureienskappe van die model se uitset. Byvoorbeeld, jy sal waarskynlik die temperatuur verhoog wanneer jy ’n toepassing skep wat kreatiewe uitsette genereer. Omgekeerd, sal jy waarskynlik die temperatuur verlaag wanneer jy ’n model bou wat prente of teks klassifiseer om sodoende die model se akkuraatheid en konsekwentheid te verbeter.	Ihyperparameter elawula iqondo lokungakhethi kwemveliso yemodeli. Amaqondo obushushu aphezulu abangela imveliso engacwangciswanga ngakumbi, ngelixa amaqondo obushushu asezantsi ephumela kwimveliso engacwangciswanga ngokuncinci. Ukukhetha obona bushushu kuxhomekeke kwiaplikheyshini ethile neempawu ezikhethiweyo zemveliso yemodeli. Umzekelo, unganyusa ubushushu xa usenza iaplikheyshini evelisa imveliso yokuyila. Ngokuchaseneyo, ubuya kuthoba iqondo lobushushu xa usakha imodeli ehlela imifanekiso okanye umbhalo ukuze kuphuculwe ukuchaneka kwemodeli nokuhambelana.	Uhlelo lokulawula ukufunda ngemishini olulawula izinga lokungavami kwemiphumela yohlelo. Umphumela wezinga eliphezulu luwumphumela ongavamile kakhulu, kube umphumela wamadszingakushisa aphansi wumphumela ongavamile kancane. Ukukhetha izingakushisa eliphuma phambili kuncike ohlweni olubaluliwe kanye nezakhi ezikhethwayo emphumeleni wohlelo. isibonelo, ungase unyuse izingakushisa uma wakha uhlelo olukhiqiza umphumela wobuciko. Kungenjalo, ungase wehlise izingakushisa lapho wakha uhlelo olubeka izithombe ngokwamazinga ukuze kuthuthukiswe ukunemba nokuhambisana kohlelo.
google_research_ai_terminology_81	True negative	ware negatief	Inegethivu eyinyani	Okungekuhle ngokweqiniso	TN / An example in which the model correctly predicts the negative class. For example, the model infers that a particular email message is not spam, and that email message really is not spam.	WN / ’n Voorbeeld waarin die model die negatiewe klas korrek voorspel. Byvoorbeeld, die model lei af dat ’n spesifieke e-posboodskap nie gemorspos is nie, en die e-pos dan werklik nie gemorspos is nie.	TN / Umzekelo apho imodeli iqikelela ngokuchanekileyo udidi olunegethivu. Umzekelo, imodeli ichaza ukuba umyalezo othile weimeyile ayisospem, ube loo myalezo weimeyile ungesospem nyhani.	i-TN / Isibonelo sokuqagela kohlelo ngokufanele ezingeni lokungekuhle. Isibonelo, uhlelo lusho ukuthi umyalezo we-imeyli othile awusiwo ugaxekile, futhi umyalezo we-imeyli awusiwo ngempela ugaxekile.
google_research_ai_terminology_82	True positive	ware positief	Iphozithivu eyinyani	Okuhle ngokweqiniso	TP / An example in which the model correctly predicts the positive class. For example, the model infers that a particular email message is spam, and that email message really is spam.	WP / ’n Voorbeeld waarin die model die positiewe klas korrek voorspel. Byvoorbeeld, die model lei af dat ’n spesifieke e-posboodskap gemorspos is, en die e-pos dan werklik gemorspos is.	TP / Umzekelo apho imodeli iqikelela ngokuchanekileyo udidi oluphozithivu. Umzekelo, imodeli ithi umyalezo othile weimeyile sispem, ibe loo myalezo weimeyile ube usisipem nyhani.	i-TP / Isibonelo sokuqagela kohlelo ngokufanele ezingeni lokuhle. Isibonelo, uhlelo lusho ukuthi umyalezo we-imeyli othile uwugaxekile, futhi umyalezo we-imeyli ngempela uwugaxekile.
google_research_ai_terminology_83	Underfitting	onderpassing	Ukungalingani okunganeno	uhlelo olubuthaka	Producing a model with poor predictive ability because the model hasn't fully captured the complexity of the training data.	Om ’n model te produseer met swak voorspellingvermoë omdat die model nie die kompleksiteit van die opleidingdata ten volle aangeteken het nie.	Ukuvelisa imodeli enesakhono esilambathayo sokuqikelela ngenxa yokuba imodeli khange ibambe ngokupheleleyo ubunzima bedatha yoqeqesho.	Ukukhiqiza uhlelo olungaqinile kwezokuqagela ngoba uhlelo longakayiqophi imininingwane yokuqeqesha enzulu.
google_research_ai_terminology_84	Undersampling	afbalansering	Isampuli enganeno	Ukususa amasampula amakhulu	Removing examples from the majority class in a class-imbalanced dataset in order to create a more balanced training set.	Die verwydering van voorbeelde van die meerderheidklas in ’n klasongebalanseerde datastel om sodoende ’n meer gebalanseerde opleidingstel te skep.	Ukususa imizekelo kwiklasi yesininzi kwiseti yedatha kwiklasi elinganayo ukwenzela ukudala isethi yoqeqesho olulungeleleneyo.	Ukususa izibonelo ezingeni lokukhulu ohlwini lwemininingwane olungalingene ngokufanele ukuze kwakhiwe uhlu lokuqeqesha olulingene.

google_research_ai_terminology_0

Accuracy

akkuraatheid

Ukuchaneka

Ukunemba

It is the proportion of all classifications that were correct, whether positive or negative. Emails example: In the spam classification example, accuracy measures the fraction of all emails correctly classified.

Dit is die gedeelte van alle klassifikasies wat korrek was, ongeag of dit positief of negatief was. E-posvoorbeeld: In die strooiposklassifikasievoorbeeld meet akkuraatheid die breukdeel van alle e-posse wat korrek geklassifiseer is.

Ngumlinganiselo wazo zonke iindidi zokuhlela ebezichanekile, nokuba bezilungile okanye zingalunganga. Umzekelo we-imeyile: Kumzekelo wokuhlelwa kwe-spam, ukuchaneka kulinganisa iqhezu lazo zonke ii-imeyile ezihlelwe ngokuchanekileyo.

Kuyisilinganiso sakho konke ukuhlunga obekunembile, kungakhathaliseki ukuthi kuhle noma kubi. Isibonelo sama-imeyili: Esibonelweni sokuhlungwa kukagaxekile, ukunemba kulinganisa ingxenye yawo wonke ama-imeyili ahlungwe ngokunembile.

google_research_ai_terminology_1

Active learning

aktiewe leer

Ukufunda ngokusebenza

Ukufunda okusebenzayo

A special case of machine learning in which a learning algorithm can interactively query a human user (or some other information source), to label new data points with the desired outputs.

’n Spesiale geval van masjienleer waarin ’n leeralgoritme interaktief navraag by ’n menslike gebruiker (of ’n ander inligtingbron) kan doen om nuwe datapunte met die verlangde uitsette te etiketteer.

Imeko ekhethekileyo yokufunda koomatshini apho i-algorithm yokufunda inokubuza ngokudibeneyo kumsebenzisi ongumntu (okanye omnye umthombo wolwazi), ukulebheyila amanqaku amatsha edatha kunye neziphumo ezifunwayo.

Isimo esikhethekile sokufunda ngomshini lapho i-algorithm yokufunda ingabuza ngokuxhumana nomsebenzisi ongumuntu (noma omunye umthombo wolwazi), ukulebula amaphoyinti edatha amasha ngemiphumela efiselekayo.

google_research_ai_terminology_2

AI alignment

KI-belyning

Ulungelelwaniso lweAI

Ukuqondanisa kwe-AI

Aims to steer AI systems toward a person's or group's intended goals, preferences, and ethical principles. An AI system is considered aligned if it advances the intended objectives. A misaligned AI system pursues unintended objectives.

Dit is daarop gemik om AI-stelsels in die rigting van ’n persoon of groep se bedoelde doelwitte, voorkeure en etiese beginsels te stuur. ’n AI-stelsel word as gerig beskou as dit die bedoelde doelwitte bevorder. ’n Wangerigte AI-stelsel streef onbedoelde doelwitte na.

Ijolise ekuqhubeni iinkqubo ze-AI ukuya kwiinjongo ezijoliswe kumntu okanye iqela, ukhetho, nemigaqo yokuziphatha. Inkqubo ye-AI ithathwa njengelungelelanisiweyo ukuba iqhubela phambili iinjongo ekujoliswe kuzo. Inkqubo ye-AI engalunganga ilandela iinjongo ezingalindelekanga.

Imigomo yokuqondisa amasistimu we-AI emigomweni ehloselwe umuntu noma iqembu, okuncanyelwayo, nemithetho yokuziphatha. Isistimu ye-AI ibhekwa njengeqondanisiwe uma ithuthukisa imigomo ehlosiwe. Isistimu ye-AI engaqondanisiwe kahle iya emigomweni engahlosiwe.

google_research_ai_terminology_3

Algorithmic bias

algoritmesydigheid

Ukuthatha icala kwealgorithm

Ukuchema ngokwe-algorithmic

A systematic and repeatable errors in a computer system that create "unfair" outcomes, such as "privileging" one category over another in ways different from the intended function of the algorithm.

’n Sistematiese en herhaalbare fout in ’n rekenaarstelsel wat “onregverdige” uitkomste skep, soos om een kategorie bo ’n ander te “bevoordeel” op maniere wat van die bedoelde funksie van die algoritme afwyk.

Iimpazamo ezicwangcisiweyo neziphindaphindwayo kwinkqubo yekhompyutha edala iziphumo "ezingalunganga", ezifana "nokubeka phambili" olunye udidi ngaphezulu kolunye ngeendlela ezahlukileyo kumsebenzi ocetyiweyo we-algorithm.

Amaphutha wesistimu naphindekayo kusistimu yekhompyutha asungula imiphumela "engafanele", "njengokuphakamisa" isigaba esithile ngaphezu kwesinye ngezindlela ezihlukile emsebenzini obuhlosiwe we-algorithm.

google_research_ai_terminology_4

Algorithmic transparency

algoritmedeursigtigheid

Ukubonakala kwealgorithm

Ukungafihlwa ngokwe-algorithmic

The principle that the factors that influence the decisions made by algorithms should be visible, or transparent, to the people who use, regulate, and are affected by systems that employ those algorithms.

Die beginsel dat die faktore wat die besluite beïnvloed wat deur algoritmes geneem word, sigbaar, of deursigtig, moet wees vir die mense wat stelsels wat daardie algoritmes benut, gebruik en reguleer en daardeur geraak word.

Umgaqo wokuba izinto ezichaphazela izigqibo ezenziwe ziialgorithm kufuneka zibonakale, okanye zibe lubala kubantu abazisebenzisayo, abazilawulayo, nabachaphazelekayo ziinkqubo ezisebenzisa ezo zixhobo.

Umthetho wokuthi izici ezinegalelo ezinqumweni ezenziwa ama-algorithms kufanele zibonakale, noma zingafihleki, kubantu abasebenzisa, baqondise, futhi bathintwe amasistimu asebenzisa lawo ma-algorithm.

google_research_ai_terminology_5

Artificial Intelligence

kunsmatige intelligensie

Ubungqondi bekhompyutha

Ubuhlakani Bokwakhiwe

AI / In its broadest sense, is intelligence exhibited by machines, particularly computer systems. It is a field of research in computer science that develops and studies methods and software that enable machines to perceive their environment and use learning and intelligence to take actions that maximize their chances of achieving defined goals.

KI / Dit is in die wydste betekenis intelligensie wat deur masjiene, veral rekenaarstelsels, getoon word. Dit is ’n navorsingsveld in rekenaarwetenskap wat studiemetodes en sagteware ontwikkel wat dit vir masjiene moontlik maak om hul omgewing waar te neem en leer en intelligensie te gebruik om handelinge uit te voer wat hul kans so groot moontlik maak om gedefinieerde doelwitte te bereik.

AI / Xa ichazwa ngeyona ndlele igabalala, bubukrelekrele obubonakaliswa ngoomatshini, ingakumbi iinkqubo zekhompyutha. Licandelo lophando kwisayensi yekhompyutha ephuhlisa ize ifunda iindlela nesoftware eyenza ukuba oomatshini babone okuwungqongileyo kwaye usebenzise ukufunda nobukrelekrele ukuthatha iintshukumo ezandisa amathuba awo okufikelela kwiinjongo ezichaziweyo.

Ngomqondo obanzi, ukuhlakanipha okubonakaliswa imishini, ikakhulu amasistimu ekhompyutha. Iyinkundla yocwaningo lwesayensi yekhompyutha esungula futhi icwaninge izindlela nesofthiwe evumela imishini ukuba ibone indawo futhi isebenzise ukufunda nokuhlakanipha ukuze ithathe izinyathelo ezikhuphula amathuba ayo okufinyelela imigomo ebekiwe.

google_research_ai_terminology_6

Artificial Neural Network

kunsmatige neurale netwerk

INethiwekhi yeNgqondo Eyenziweyo

Ubuxhakaxhaka Bokunikezelana Obakhiwe

ANN / In machine learning, a neural network (also artificial neural network or neural net, abbreviated ANN or NN) is a model inspired by the structure and function of biological neural networks in animal brains. An ANN consists of connected units or nodes called artificial neurons, which loosely model the neurons in the brain.

KNN / In masjienleer is ’n neurale netwerk (ook kunsmatige neurale netwerk of neurale net, afgekort as KNN of NN) ’n model wat geïnspireer is deur die struktuur en funksie van biologiese neurale netwerke in die breine van diere. ’n KNN bestaan uit gekoppelde eenhede of nodusse genaamd kunsmatige neurone, wat losweg na die neurone in die brein gemodelleer is.

ANN / Ekufundweni koomatshini, i-neural nethwekhi (ne-artificial neural nethwekhi okanye i-neural net, i-ANN xa ifinyeziwe okanye i-NN) yimodeli engumzekelo ekhuthazwa sisakhiwo nomsebenzi we-biological neural networks kubuchopho bezilwanyana. I-ANN iqulathe iiyunithi ezidityanisiweyo okanye iindawo ezibizwa ngokuba zii-neuron ezenziweyo, ezibonisa ngokukhululekileyo imithambo-luvo ebuchotsheni.

Ekufundeni komshini, ineural network (ebizwa nangokuthi yi-artificial neural network noma ineural net, enesifinyezo esithi ANN noma NN) iyimodeli esekelwa ukwakheka nokusebenza kwamabiological neural network ebuchosheni bezilwane. I-ANN iqukethe amayunithi noma amanodi axhumene abizwa ngokuthi ama-artificial neuron, alingisa amaneuron asebuchosheni.

google_research_ai_terminology_7

Autonomous Vehicles

outonome voertuig

Izithuthi ezizilawulayo

Izimoto Ezizishayelayo

Refers to a car that is capable of operating with reduced or no human input. Synonyms: self-driving car, driverless car, robotaxi, robotic car or robo-car.

Dit verwys na ’n motor wat in staat is om met verminderde of geen menslike insette nie te werk. Sinonieme: selfbestuurmotor, bestuurderlose motor, robottaxi, robotiese motor of robotmotor.

Ibhekisa kwimoto ekwaziyo ukusebenza ngegalelo lomntu elinciphileyo okanye kungekho galelo lomntu. Izithethantonye: imoto eziqhubayo, imoto engenamqhubi, i-robotaxi, imoto yerobhothi okanye i-robo-car.

Zibhekisela emotweni ekwazi ukuhamba ngokuqondisa okuncane komuntu noma ngaphandle kwako. Amagama afanayo: imoto ezishayelayo, imoto engenamshayeli, irobotaxi, imoto eyirobhothi noma irobo-car.

google_research_ai_terminology_8

Bias-variance tradeoff

sydigheid-variansie-afruiling

Ukuthatha icala kwiyantlukwano yorhwebo

Ilumbanisamaphutha

Describes the relationship between a model's complexity, the accuracy of its predictions, and how well it can make predictions on previously unseen data that were not used to train the model.

Dit beskryf die verhouding tussen ’n model se kompleksiteit, die akkuraatheid van die voorspellings daarvan, en hoe goed dit voorspellings kan doen op grond van data wat nie vroeër gesien is nie en nie gebruik is om die model op te lei nie.

Ichaza unxulumano phakathi kokuntsokotha kwemodeli, ukuchaneka koqikelelo lwayo, nendlela enokwenza ngayo uqikelelo kwidatha ebingabonwa ngaphambili engazange isetyenziswe ekuqeqesheni imodeli.

Ichaza ubuhlobo phakathi kobunkimbinkimbi bemodeli, ukunemba kokuqagela kwayo, nendlela ekwazi ukuqagela kahle ngayo kudatha engakaze ibonwe ngaphambilini engazange isetshenziswe ekuqeqesheni leyo modeli.

google_research_ai_terminology_9

Chatbot

kletsbot

Ichatbot

Ifuzelankulumo

A software application or web interface that is designed to mimic human conversation through text or voice interactions. Also spelled as chat bot.

’n Sagtewareapp of webkoppelvlak wat ontwerp is om menslike gesprek deur teks- of steminteraksies na te boots. In Engels gespel as “chatbot” of “chat bot”.

Iaplikheyshini yesoftware okanye ujongano lwewebhu eyilelwe ukulinganisa incoko yabantu ngokubhaliweyo okanye ukusetyenziswa kwelizwi. Ikwapelwe ngokuthi chat bot.

I-application yesofthiwe noma isixhumi esibonakalayo sewebhu esiklanyelwe ukulingisa ingxoxo yabantu ngombhalo noma ngokuxhumana ngezwi. Iphinde ibhalwe nangokuthi chat bot.

google_research_ai_terminology_10

Computer vision

rekenaarvisie

Umbono wekhompyutha

An interdisciplinary field that deals with how computers can be made to gain high-level understanding from digital images or videos. From the perspective of engineering, it seeks to automate tasks that the human visual system can do. As a scientific discipline, computer vision is concerned with the theory behind artificial systems that extract information from images.

’n Interdissiplinêre vakgebied wat gaan oor hoe rekenaars gemaak kan word om hoëvlakbegrip van digitale prente of video’s te kry. Uit ’n ingenieursoogpunt is dit daarop gemik om take te outomatiseer wat die menslike visuele stelsel kan verrig. As ’n wetenskaplike dissipline is rekenaarvisie gemoeid met die teorie agter kunsmatige stelsels wat inligting uit prente haal.

Inkalo yezifundo ezahlukeneyo ejongene nendlela iikhompyutha ezinokuthi zenziwe ngayo ukufumana ukuqonda okuphezulu kwimifanekiso yedijithali okanye iividiyo. Ngokwembono yobunjineli, ifuna ukuzenzela imisebenzi enokwenziwa yinkqubo yokubona yomntu. Njengoqeqesho lwezenzululwazi, umbono wekhompyutha ujongene nethiyori esemva kweenkqubo ezenziweyo ezikhupha inkcazelo kwimifanekiso.

Inkundla eyizinhlobonhlobo egxile endleleni okungenziwa ngayo ukuba amakhompyutha athole ukuqonda kwezinga eliphezulu ngezithombe namavidiyo edijithali. Ngokombono wonjiniyela, ihlose ukwenza imisebenzi eyenziwa isistimu yokubuka komuntu. Ngokomthetho wesayensi, ukubona kwekhompyutha kukhathazeke ngendaba yamasistimu okwenziwa adonsa imininingwane ezithombeni.

google_research_ai_terminology_11

Conversational AI

gesprek-KI

I-AI yokuncokola

Ingxoxo yobuhlakani obakhiwe

A type of artificial intelligence (AI) that can simulate human conversation thanks to natural language processing (NLP) technologies.

’n Soort kunsmatige intelligensie (AI) wat menslike gesprekke kan naboots danksy tegnologie vir natuurlike taalverwerking (NTV).

Uhlobo lwe-artificial intelligence (AI) ekwazi ukulinganisa incoko yabantu ngoncedo lobuchwephesha bokulungiswa kolwimi lwendalo (NLP).

Uhlobo lobuchwepheshe bokwakhiwe (AI) olungalingisa ingxoxo yabantu ngenxa yobuchwepheshe bokucubungula kolimi lwemvelo (NLP).

google_research_ai_terminology_12

Data anonymization

data-anonimisering

Ukwenza idatha ifihlwe iinkcukacha

Ufihlomvelaphi lwemininingo

A process by which personal data is altered in such a way that a data subject can no longer be identified directly or indirectly, either by the data controller alone or in collaboration with any other party.

’n Proses waardeur persoonlike data só gewysig word dat ’n datasubjek nie meer direk of indirek deur óf die databeheerder alleen, óf in samewerking met enige ander party geïdentifiseer kan word nie.

Inkqubo apho idatha yomntu iguqulwa khona ngendlela yokuba umxholo wedatha ungasakwazi ukuchongwa ngokuthe ngqo okanye ngokungathanga ngqo, nokuba ngumlawuli wedatha yedwa okanye ngokubambisana nalo naliphi na elinye iqela.

Inqubo yokuguqulwa kwedatha yomuntu ngendlela yokuthi umnikazi wedatha angabe esakwazi ukutholwa ngokuqondile noma ngokungaqondile, kungaba ngesilawuli sedatha sisodwa noma sihlangene nanoma yikuphi okunye.

google_research_ai_terminology_13

Data augmentation

data-aanvulling

Ukwandiswa kwedatha

Isandisamininingo

A statistical technique which allows maximum likelihood estimation from incomplete data.

’n Statistiese tegniek wat ’n skatting van die grootste waarskynlikheid uit onvolledige data moontlik maak.

Ubuchule beenkcukacha-manani obuvumela uqikelelo lokunokwenzeka okukhulu okusuka kwidatha engaphelelanga.

Ikhono lezibalo elivumela ukulinganisa kokuqagela okukhulu ngedatha engaphelele.

google_research_ai_terminology_14

Data de-anonymization

datadeanonimisering

Ukwaziwa kwedatha yaziwe kwakhona

Ivezamvelaphi lemininingo

It is the practice of matching anonymous data (also known as de-identified data) with publicly available information, or auxiliary data, in order to discover the person the data belong to. Also known as 'data re-identification'.

Dit is die praktyk om anonieme data (ook bekend as gedeïdentifiseerde data) te pas by inligting wat publiek beskikbaar is, of aanvullende data, om uit te vind aan watter persoon die data behoort. Dit is ook bekend as “dataheridentifisering”.

Isenzo sokuthelekisa idatha engachazwanga (ekwabizwa ngokuba yidatha eyekisiweyo ukuchonga) enenkcukacha ezifumaneka esidlangalaleni, okanye iinkcukacha ezincedisayo, ukuze kufunyanwe ukuba ngubani umntu wale datha. Ikwabizwa ngokuba 'kukuphinda kuchongwe idatha'.

Inqubo yokuqhathanisa idatha engaziwa (eyaziwa nangokuthi yide-identified data) enemininingwane etholakala esidlangalaleni, noma idatha eseceleni, ukuze kutholakale umuntu ongumnikazi waleyo datha. Yaziwa nangokuthi 'yidata re-identification'.

google_research_ai_terminology_15

Data mining

dataontginning

Ukwembiwa kweenkcukacha

Ukuhlwaya imininingo

An interdisciplinary subfield of computer science and statistics with an overall goal of extracting information (with intelligent methods) from a data set and transforming the information into a comprehensible structure for further use.

’n Interdissiplinêre subvakgebied van rekenaarwetenskap en statistiek waarvan die oorkoepelende doel is om inligting (met intelligente metodes) uit ’n datastel te haal en die inligting tot ’n verstaanbare struktuur te omskep vir verdere gebruik.

Icandelo elingaphantsi kwezifundo zesayensi yekhompyutha neenkcukacha-manani enenjongo yokukhupha inkcazelo (ngeendlela ezikrelekrele) ukusuka kwiseti yedatha nokuguqula inkcazelo kwisakhiwo esiqondakalayo ukuze sisetyenziswe ngakumbi.

Inkundla encane eyizinhlobonhlobo yesayensi yekhompyutha nezibalo enomgomo wokudonsa imininingwane (ngezindlela ezihlakaniphile) kusethi yedatha nokuguqula leyo mininingwane ibe yisakhiwo esiqondakalayo ukuze isetshenziswe.

google_research_ai_terminology_16

Deep Learning

diepleer

Ukufunda nzulu

Ukufunda usebenzisa izici

Deep learning refers to a class of machine learning algorithms in which a hierarchy of layers is used to transform input data into a slightly more abstract and composite representation. For example, in an image recognition model, the raw input may be an image (represented as a tensor of pixels). The first representational layer may attempt to identify basic shapes such as lines and circles, the second layer may compose and encode arrangements of edges, the third layer may encode a nose and eyes, and the fourth layer may recognize that the image contains a face.

Diepleer verwys na ’n klas masjienleeralgoritmes waarin ’n hiërargie lae gebruik word om invoerdata tot ’n effens meer abstrakte en saamgestelde verteenwoordiging te omskep. Byvoorbeeld, in ’n prentherkenningmodel kan die onverwerkte invoer ’n prent wees (wat as ’n tensor pieksels verteenwoordig word). Die eerste verteenwoordigende laag kan probeer om basiese vorms, soos lyne en sirkels, te identifiseer; die tweede laag kan rangskikkings van rande saamstel en enkodeer; die derde laag kan neuse en oë enkodeer; en die vierde laag kan herken dat die prent ’n gesig bevat.

Ukufunda nzulu kubhekisele kwiklasi yokufunda koomatshini beealgorithm apho ukulandelelana kweeleya kusetyenziselwa ukuguqula idatha efakiweyo ibe yinto engabonakaliyo nomboniso odibeneyo. Umzekelo, kwimodeli yokuqaphela umfanekiso, umthombo wokufakwayo unokuba ngumfanekiso (omelwe njengetensor yeepikseli). Ileya yokuqala omeleyo unokuzama ukuchonga iimilo ezisisiseko ezifana nemigca kunye nezangqa, ileya yesibini inokuqamba ize idibanise ulungelelwaniso lwemiphetho, ileya yesithathu inokubhaqa impumlo namehlo, ize ileya yesine iqonde ukuba umfanekiso uqulathe ubuso.

Ideep learning ibhekisela esigabeni sama-algorithm okufunda komshini lapho amazinga ezendlalelo esetshenziselwa ukuguqula idatha efakwayo ibe yisimo esifihlekile nesididiyelwe. Ngokwesibonelo, kumodeli yokuqashelwa kwesithombe, okufakwayo okungahlungiwe kungase kube isithombe (esimelelwa yitensor yamaphikseli). Isendlalelo sokuqala esimelelayo singase sizame ukubona izinhlobo zezinto ezivamile njengolayini neziyingi, isendlalelo sesibili singase sibhale futhi sifake ikhodi ekuhlelweni kwamachopho, isendlalelo sesithathu singase sifake ikhodi yekhala namehlo, kanti isendlalelo sesine singase sibone ukuthi isithombe siqukethe ubuso.

google_research_ai_terminology_17

Deepfake

diepvervalsing

Ideepfake

Ifihlabuwuwa

Images, videos, or audio which are edited or generated using artificial intelligence tools, and which may depict real or non-existent people. They are a type of synthetic media.

Prente, video’s of oudio wat met gebruik van nutsgoed met kunsmatige intelligensie geredigeer of gegenereer is, en wat regte of niebestaande mense kan uitbeeld. Dit is ’n soort sintetiese media.

Imifanekiso, iividiyo, okanye okumanyelwayo okuedithiweyo okanye okuveliswa kusetyenziswa izixhobo zobukrelekrele ezenziweyo, ezinokuthi zimele abantu bokwenyani okanye abangekhoyo. Ziluhlobo lwemidiya eyenziweyo.

Izithombe, amavidiyo, noma okulalelwayo okuhlelwa noma kukhiqizwe ngamathuluzi wobuchwepheshe bokwakhiwe, futhi okungabonisa abantu bangempela noma abangekho ngokoqobo. Kuyizinhlobo zesynthetic media.

google_research_ai_terminology_18

Explainable AI

verklaarbare KI

I-AI ecacisekayo

Ubuhlakani Obakhiwe Obuchazekayo

XAI / Either refers to an artificial intelligence (AI) system over which it is possible for humans to retain intellectual oversight, or refers to the methods to achieve this.

VKI / Dit verwys óf na ’n stelsel met kunsmatige intelligensie (AI) waaroor dit vir mense moontlik is om intellektuele toesig te behou, óf dit verwys na die metodes om dit reg te kry.

XAI / Ibhekisa kwinkqubo yobukrelekrele bokwenziwa (AI) apho kunokwenzeka ukuba abantu bagcine ukongamela ubukrelekrele, okanye ibhekisa kwiindlela zokwazi ukuphumeza oku.

Ibhekisela kusistimu yobuchwepheshe bokwakhiwe (AI) eyenza kwenzeke ukuthi abantu baqhubeke beqondisa ngokuhlakanipha, noma kubhekisela ezindleleni zokukufinyelela lokhu.

google_research_ai_terminology_19

Fairness

regverdigheid

Ubulungisa

Ukungachemi

in AI / Fairness in machine learning refers to the various attempts at correcting algorithmic bias in automated decision processes based on machine learning models. Decisions made by computers after a machine-learning process may be considered unfair if they were based on variables considered sensitive. For example gender, ethnicity, sexual orientation or disability.

in KI / Regverdigheid in masjienleer verwys na die verskillende pogings om algoritmiese vooroordeel in geoutomatiseerde besluitnemingprosesse op grond van masjienleermodelle reg te stel. Besluite wat ná ’n masjienleerproses deur rekenaars geneem word, kan as onregverdig beskou word as hulle gegrond is op veranderlikes wat as sensitief beskou word. Byvoorbeeld gender, etnisiteit, seksuele oriëntasie of gestremdheid.

kwi-AI / I-Fairness ekufundweni koomatshini ibhekiselele kwiinzame ezahlukeneyo zokulungisa ukuthatha icala kweealgorithm kwiinkqubo zezigqibo ezizenzekelayo ezisekelwe kwiimodeli zokufundwa koomatshini. Izigqibo ezenziwe ziikhompyutha emva kwenkqubo yokufunda koomatshini zinokujongwa njengezingafanelekanga ukuba bezisekelwe kwizinto ezithathwa njengezinovakalelo. Umzekelo isini, ubuhlanga, ukuziqhelanisa ngesondo okanye ukukhubazeka.

ku-AI / Ukungachemi ekufundeni komshini kubhekisela emizamweni ehlukahlukene yokulungisa ukuchema kwama-algorithm ezinqubweni zezinqumo ezizenzekelayo ezisekelwe kumamodeli okufunda komshini. Izinqumo ezenziwa amakhompyutha ngemva kwenqubo yokufunda komshini zingase zibhekwe njengezingafanele uma zisekelwe ezingxenyekazini ezibhekwa njengezizwelayo. Ngokwesibonelo ubulili, ubuzwe, ubulili oyamaniseka nabo noma ukukhubazeka.

google_research_ai_terminology_20

Few-shot prompting

veelvoorbeeldinstruksieteks

Uyalelo olumizekelo mininzi

Ukuyalela ngokumbalwa

A prompting strategy that includes two or more examples of the desired input and output

’n Porboodskapstrategie wat twee of meer voorbeelde van die verlangde invoer en uitvoer insluit

Isicwangciso esiyalelayo esibandakanya imizekelo emibini okanye ngaphezulu yokufakwayo okufunekayo nesiphumo

Isu leprompting oluhlanganisa izibonelo ezimbili noma ngaphezulu zokufakwayo nomphumela ofunwayo

google_research_ai_terminology_21

Fine-tuning

verfyning

Ukutyhuna kakuhle

Ukucolisisa

In deep learning, fine-tuning is an approach to transfer learning in which the parameters of a pre-trained model are trained on new data.

In diepleer is fyn instelling ’n benadering om leer oor te dra waarin die parameters van ’n voorafopgeleide model met nuwe data opgelei word.

Ekufundeni nzulu, i-fine-tuning yindlela yokudlulisa ukufunda apho iipharamitha zemodeli eqeqeshwe kwangaphambili ziqeqeshelwe idatha entsha.

Ekufundeni ngokujulile, ifine-tuning iyindlela yokudlulisela ukufunda lapho amapharamitha wemodeli eqeqeshwe kusengaphambili eqeqeshelwa idatha entsha.

google_research_ai_terminology_22

Generative AI

generatiewe KI

I-AI yokuvelisa

Ubuhlakani Obakhiwe Obukhiqizayo

Artificial intelligence capable of generating text, images, videos, or other data using generative models, often in response to prompts. Generative AI models learn the patterns and structure of their input training data and then generate new data that has similar characteristics.

Kunsmatige intelligensie wat die vermoë het om teks, prente, video’s of ander data met gebruik van generatiewe modelle te genereer, dikwels in reaksie op porboodskappe. Generatiewe-AI-modelle leer die patrone en struktuur van hul invoeropleidingdata en genereer dan nuwe data wat soortgelyke kenmerke het.

I-artificial intelligence ekwazi ukuvelisa umbhalo, imifanekiso, iividiyo, okanye enye idatha isebenzisa imodeli yokuzivelisela, ehlala iphendula imiyalelo. Iimodeli ze-AI Yokuvelisa zifunda iiphatheni nolwakhiwo lwedatha yoqeqesho lwazo zize emva koko zivelise idatha entsha eneempawu ezifanayo.

Ubuchwepheshe bokwakhiwe bukwazi ukukhiqiza umbhalo, izithombe, amavidiyo, nenye idatha ngamamodeli akhiqizayo, ngokuvamile ngokusabela emiyalelweni. Amamodeli we-AI ekhiqizayo afunda amaphethini nesakhiwo sedatha yokuqeqeshwa yokufakwayo bese ekhiqiza idatha entsha enezici ezifanayo.

google_research_ai_terminology_23

Hallucination

hallusinasie

Ukubona izinto ezingekhoyo

Ukudideka

A response generated by AI which contains false or misleading information presented as fact.

’n Antwoord gegenereer deur AI wat vals of misleidende inligting bevat, maar as ’n feit voorgehou word.

Impendulo eveliswa yi-AI equlethe inkcazelo engeyonyani okanye elahlekisayo eboniswe njengenyaniso.

Impendulo ekhiqizwa yi-AI equkethe imininingwane engamanga noma edukisayo enikezwa njengeyiqiniso.

google_research_ai_terminology_24

Human-in-the-Loop

mens-wat-ingesluit-word

Ihuman-in-the-Loop

Inxumano Yomuntu Nomshini

HITL / In machine learning, HITL is used in the sense of humans aiding the computer in making the correct decisions in building a model.

MWIW / In masjienleer word MWIW gebruik in die sin van mense wat die rekenaar bystaan om die korrekte besluite te neem wanneer ’n model gebou word.

HITL / Ekufundeni koomatshini, i-HITL isetyenziswa kwingqiqo yabantu encedisa ikhompyutha ekwenzeni izigqibo ezichanekileyo ekwakheni imodeli.

Ekufundeni komshini, i-HITL isetshenziswa ngomqondo wokuthi abantu basiza ikhompyutha ekwenzeni izinqumo ezinembile ekwakheni imodeli.

google_research_ai_terminology_25

Large Language Model

groot taalmodel

IModeli yoLwimi eNkulu

Insuselakuyo Yolimi Ngokubanzi

LLM / A computational model capable of language generation or other natural language processing tasks. As language models, LLMs acquire these abilities by learning statistical relationships from vast amounts of text during a self-supervised and semi-supervised training process.

GTM / ’n Rekenaarmodel wat taal kan genereer of ander take met natuurlike taalverwerking kan uitvoer. As taalmodelle verkry GTM’e hierdie vermoëns deur statistiese verhoudings uit ontsaglike hoeveelhede teks te leer tydens ’n opleidingproses onder selftoesig of semitoesig.

LLM / Imodeli yekhompyutha ekwaziyo ukuvelisa ulwimi okanye eminye imisebenzi yokulungisa ulwimi lwendalo. Njengeemodeli zolwimi, ii-LLM zifumana obu buchule ngokufunda unxulumano lwamanani ukusuka kumthamo omkhulu wezibhalo ngexesha lenkqubo yoqeqesho oluzigadelayo noluphantsi kweliso eliphakathi.

Imodeli yekhompyutha ekwazi ukukhiqiza ulimi noma eminye imisebenzi yokucubungula kolimi lwemvelo. Njengamamodeli wolimi, ama-LLM athola lawa makhono ngokufunda ubuhlobo bezibalo emananini amaningi ombhalo phakathi nenqubo yokuqeqesha ozigada ngokwakho kuyo negadwa ngokwengxenye.

google_research_ai_terminology_26

Machine Learning

masjienleer

IMachine Learning

Ukufunda Komshini

ML / A field of study in artificial intelligence concerned with the development and study of statistical algorithms that can learn from data and generalize to unseen data and thus perform tasks without explicit instructions.

ML / ’n Studierigting in kunsmatige intelligensie wat ingestel is op die ontwikkeling en bestudering van statistiese algoritmes wat uit data kan leer en dit vir onbekende data kan veralgemeen, en gevolglik take sonder uitdruklike instruksies kan uitvoer.

ML / Icandelo lokufunda kwi-artificial intelligence ejongene nophuhliso nokufunda ii-algorithim zamanani anokuthi afunde kwidatha kwaye enze ngokubanzi kwidatha engabonakaliyo kwaye ngaloo ndlela enze imisebenzi ngaphandle kwemiyalelo ecacileyo.

Inkundla yokucwaninga ebuchwephesheni bokwakhiwe egxile ekwakhekeni nasocwaningweni lwama-algorithm ezibalo engafunda kudatha futhi iyiguqulele kudatha engabonakali ngaleyo ndlela yenze imisebenzi ngaphandle kweziqondiso ezicacile.

google_research_ai_terminology_27

Multimodal learning

multimodale leer

Ukufunda ngeeModali ezininzi

Ukufunda ngezinsuselakuyo eziningi

A type of deep learning using multiple modalities of data, such as text, audio, or images.

’n Soort diepleer wat veelvuldige modaliteite data, soos teks, oudio of prente, gebruik.

Uhlobo lokufunda olunzulu usebenzisa iindela ezininzi zedatha, ezinjengemibhalo, okumanyelwayo, okanye imifanekiso.

Uhlobo lokufunda okujulile olusebenzisa amamodali wedatha angaphezu kweyodwa, njengombhalo, okulalelwayo, noma izithombe.

google_research_ai_terminology_28

Natural Language Processing

natuurliketaalverwerking

UkuSetyenzwa koLwimi lweNdalo

Ukucubungula Kolimi Lwemvelo

NLP / An interdisciplinary subfield of computer science and artificial intelligence, primarily concerned with providing computers with the ability to process data encoded in natural language. Major tasks in natural language processing are speech recognition, text classification, natural-language understanding, and natural-language generation.

NTV / ’n Interdissiplinêre subvakgebied in rekenaarwetenskap en kunsmatige intelligensie wat hoofsaaklik daarop ingestel is om aan rekenaars die vermoë te gee om data te verwerk wat in natuurlike taal geënkodeer is. Belangrike take in natuurlike taalverwerking is spraakherkenning, teksklassifikasie, natuurlike taalbegrip en natuurlike taalgenerering.

NLP / Indawo engaphantsi kwezifundo zenzululwazi yekhompyutha ne-artificial intelligence, yeyona nto ijongene nokubonelela iikhompyutha ngesakhono sokucubungula idatha efakwe ngekhowudi ngolwimi lwendalo. Imisebenzi ephambili ekusetyenzweni kolwimi lwendalo kukunakanwa kwentetho, ukuhlelwa kwemibhalo, ukuqonda ulwimi lwendalo, nokuveliswa kolwimi lwendalo.

NLP / Inkundla engaphezu kweyodwa yesayensi yekhompyutha nobuchwepheshe bokwakhiwe, egxile ngokuyinhloko ekunikezeni amakhompyutha ikhono lokucubungula idatha ebhalwe ngekhodi yolimi olungokwemvelo. Imisebenzi emikhulu ekucubunguleni kolimi lwemvelo ukubonwa kwengxoxo, ukuhlunga umbhalo, ukuqonda ulimi lwemvelo, nokukhiqiza ulimi lwemvelo.

google_research_ai_terminology_29

Noisy data

raserige data

Idatha eninzi engelo ncedo

Imininingo enolwazi olungadingeki

Data with a large amount of additional meaningless information in it called noise.

Data met ’n groot hoeveelheid bykomende, betekenislose inligting daarin wat geraas genoem word.

Idatha enesixa esikhulu senkcazelo eyongezelelweyo engenantsingiselo kuyo ibizwa ngokuba yi-noise.

Idatha enenani elikhulu lemininingwane engenancazelo kuyo ebizwa ngomsindo.

google_research_ai_terminology_30

One-shot prompting

eenvoorbeeldinstruksieteks

Uyalelo olumzekelo mnye

Ukuyalelela ngokukodwa

A prompting strategy that includes one example of the desired input and output

’n Porboodskapstrategie wat een voorbeeld van die verlangde invoer en uitvoer insluit

Isicwangciso sokuyalela esibandakanya umzekelo omnye wento efakwayo efunekayo nesiphumo

Isu lomyalelo elibandakanya isibonelo esisodwa sokufakwayo nomphumela ofiselekayo

google_research_ai_terminology_31

Open source

oopbron

Umthombo ovulelekileyo

Umthombo ovulekile

A source code (a plain text computer program written in a programming language) that is made freely available for possible modification and redistribution.

’n Bronkode (’n rekenaarprogram in skoonteks wat in ’n programmeertaal geskryf is) wat vrylik beskikbaar gestel word vir moontlike wysiging en herverspreiding.

Ikhowudi yomthombo (inkqubo yekhompyutha ebhaliweyo ecacileyo ebhalwe ngolwimi lwenkqubo) eyenziwe ifumaneke ngokukhululekileyo ukuze ikwazi ukuguqulwa nokusasazwa kwakhona.

Ikhodi yomthombo (uhlelo lwekhompyutha lombhalo ongenalutho olubhalwe ngolimi lweprogramming) etholakala mahhala ukuze iguqulwe futhi isakazwe kabusha.

google_research_ai_terminology_32

Output

afvoer

Imveliso

Umkhiqizo

The information or creative work that an AI tool produces after it is prompted, such as an answer to a question, a text summary, an image, or a piece of music

Die inligting of kreatiewe werk wat ’n AI-nutsding lewer nadat dit gepor is, soos ’n antwoord op ’n vraag, ’n teksopsomming, ’n prent of ’n stuk musiek

Inkcazelo okanye umsebenzi wokudala oveliswa sisixhobo se-AI emva kokuyalelwa, njengempendulo yombuzo, isishwankathelo sombhalo, umfanekiso, okanye isiqwenga somculo

Imininingwane noma umsebenzi wobuciko okhiqizwa yithuluzi le-AI ngemva komyalelo, njengempendulo embuzweni, isifinyezo sombhalo, isithombe, noma umculo

google_research_ai_terminology_33

Prompt

instruksieteks

Umyalelo

An instruction, usually expressed in natural language, used to get the desired output from a generative AI engine.

’n Instruksie, gewoonlik uitgedruk in natuurlike taal, wat gebruik word om die verlangde uitvoer van ’n generatiewe-AI-enjin te kry.

Umyalelo, odla ngokuchazwa ngolwimi lwendalo, osetyenziselwa ukufumana imveliso efunekayo kwi-injini ye-AI yokuvelisa.

Isiqondiso, ngokuvamile esivezwa ngolimi olungokwemvelo, esisetshenziselwa ukuthola umphumela ofiswayo enjinini ye-AI ekhiqizayo.

google_research_ai_terminology_34

Prompt engineering

instruksieteksontwerp

Ubunjinelimyalelo

Ubunjiniyela bemiyalelo

The process of structuring an instruction that can be interpreted and understood by a generative AI model.

Die proses om ’n instruksie te struktureer wat deur ’n generatiewe-AI-model geïnterpreteer en verstaan kan word.

Inkqubo yokumisela umyalelo onokutolikwa uze uqondwe yimodeli ye-AI yokuvelisa.

Inqubo yokwakha umyalelo ongase uchazwe futhi uqondwe yimodeli ye-AI ekhiqizayo.

google_research_ai_terminology_35

Pseudonymization

pseudonimisering

IPseudonymization

Ukuzivezambumbulu

The process of obscuring data with the ability to re-identify it later. Personally identifiable information fields within a data record are replaced by one or more artificial identifiers, or pseudonyms.

Die proses om data te verdoesel met die vermoë om dit later te heridentifiseer. Velde met persoonlik identifiseerbare inligting in ’n datarekord word vervang met een of meer kunsmatige identifiseerders, of pseudonieme.

Inkqubo yokufihla idatha nokwazi ukuyichonga kwakhona kamva. Indawo yenkcazelo echongiweyo ngokobuqu kwirekhodi yedatha ethathelwe indawo sisifanisi esinye, okanye ngapezulu, okanye ii-pseudonym.

Inqubo yokufihla idatha ngendlela yokukwazi ukuyibona futhi kamuva. Izinkundla zolwazi olukhomba umuntu siqu (PII) erekhodini ledatha ziyakhishwa kufakwe izinkomba zokwenziwa ezizodwa noma ngaphezulu, noma amapseudonyms.

google_research_ai_terminology_36

Raw data

onverwerkte data

Idatha ekrwada

Imininingo engahlungiwe

Also known as primary data, are data (e.g., numbers, instrument readings, figures, etc.) collected from a source, prior to cleaning or analysis.

Ook bekend as primêre data – dit is data (byvoorbeeld nommers, instrumentlesings, syfers, ensovoorts) wat van ’n bron af ingesamel is, voordat dit opgeruim of ontleed is.

Ekwabizwa ngokuba yidatha yangaphambili, yidatha (umzekelo, amanani, ukufundwa kwezixhobo, amanani, njalo-njalo.) eqokelelwe kumthombo ngaphambi kokucoca okanye uhlalutyo.

Yaziwa nangokuthi idatha eyinhloko, iyidatha (isib., izinombolo, ukufundwa kwethuluzi, izinombolo, njll.) eqongelelwa emthonjeni, ngaphambi kokuhlanza noma ukuhlaziya.

google_research_ai_terminology_37

Reinforcement Learning

versterkingleer

Ukufunda ngokuKhuthazwa

Ukufunda Ukuthatha Izinqumo

RL / An interdisciplinary area of machine learning and optimal control concerned with how an intelligent agent ought to take actions in a dynamic environment in order to maximize the cumulative reward.

VL / ’n Interdissiplinêre terrein van masjienleer en optimale beheer wat gemoeid is met hoe ’n intelligente agent in ’n dinamiese omgewing behoort op te tree om die kumulatiewe beloning so groot moontlik te maak.

RL / Icandelo lokufunda koomatshini nolawulo olufanelekileyo olunenkxalabo ngendlela iarhente ekrelekrele ekufuneka ithathe ngayo amanyathelo kwindawo eguqukayo ukuze kwandiswe umvuzo owongezelekayo.

Indawo engaphezu kweyodwa yokufunda komshini nokulawula okuthuthukisiwe okugxile endleleni i-ejenti ehlakaniphile okufanele ithathe ngayo izinyathelo endaweni eshintshashintshayo ukuze kuthuthukiswe umvuzo oqongelelwe.

google_research_ai_terminology_38

Robotics

robotika

Irobhothiksi

IRobotiksi

The interdisciplinary study and practice of the design, construction, operation, and use of robots (a machine—especially one programmable by a computer—capable of carrying out a complex series of actions automatically).

Die interdissiplinêre bestudering en toepassing van die ontwerp, konstruksie, bedryf en gebruik van robotte (’n masjien – veral een wat deur ’n rekenaar geprogrammeer kan word – wat in staat is om ’n komplekse reeks handelinge outomaties uit te voer).

Uphononongo lwezifundo ezahlukahlukeneyo nokusebenza koyilo, ukwakhiwa, ukusebenza, kunye nokusetyenziswa kweerobhothi (umatshini–ngakumbi omnye ocwangciswayo yikhompyutha–ekwaziyo ukwenza uluhlu olunzima lwezenzo ngokuzenzekelayo).

Ucwaningo olungaphezu kolulodwa nomsebenzi wedizayini, ukwakha, ukusebenza, nokusetshenziswa kwamarobhothi (umshini—ikakhulu ongahlelwa yikhompyutha—okwazi ukufeza uchungechunge lwezenzo eziyinkimbinkimbi ngokuzenzekelayo).

google_research_ai_terminology_39

Speech recognition

spraakherkenning

Ukunakanwa kwentetho

Inzwagezwi

An interdisciplinary subfield of computer science and computational linguistics that develops methodologies and technologies that enable the recognition and translation of spoken language into text by computers.

’n Interdissiplinêre subvakgebied van rekenaarwetenskap en rekenaarlinguistiek wat metodologieë en tegnologieë ontwikkel wat die herkenning en vertaling van gesproke taal in teks deur rekenaars moontlik maak.

Icandelo elingaphantsi kwezifundo ezahlukeneyo zenzululwazi yekhompyutha neelwimi zekhompyutha eziphuhlisa iindlela nobuchwepheshe obuvumela ukuqondwa nokuguqulelwa kolwimi oluthethwayo kwisicatshulwa ngeekhompyutha.

Inkundla engaphezu kweyodwa yesayensi yekhompyutha nolimi lwekhompyutha olusungula izindlela nobuchwepheshe obuvumela ukuqashelwa nokuhunyushwa kolimi olukhulunywayo ngombhalo emakhompyutheni.

google_research_ai_terminology_40

Supervised learning

leer onder toesig

Ukufunda okujongiweyo

Ukufunda okuqashelwe

A paradigm in machine learning where input objects and a desired output value train a model.

’n Paradigma in masjienleer waar invoervoorwerpe en ’n verlangde uitvoerwaarde ’n model oplei.

I-paradigm ekufundeni koomatshini apho izinto ezifakwayo nexabiso elifunekayo lemveliso liqeqesha imodeli.

Isibonelo sokufunda komshini lapho izinto ezifakwayo nenani lomphumela ofiswayo ziqeqesha imodeli.

google_research_ai_terminology_41

Synthetic media

sintetiese media

Imidiya yokwenziwayo

Ezokwazisa ezikhiqizwa umshini

A catch-all term for the artificial production, manipulation, and modification of data and media by automated means, especially through the use of artificial intelligence algorithms, such as for the purpose of misleading people or changing an original meaning. Also known as AI-generated media.

’n Versamelterm vir die kunsmatige produksie, manipulasie en modifikasie van data en media deur geoutomatiseerde metodes, veral deur die gebruik van algoritmes met kunsmatige intelligensie, soos met die doel om mense te mislei of om ’n oorspronklike betekenis te verander. Ook bekend as AI-gegenereerde media.

Intetho ehlala isetyenziswa yemveliso eyenziweyo , ukwenziwa nokuguqulwa kwedatha kunye nemidiya ngeendlela ezizenzekelayo, ngokukodwa ngokusetyenziswa kwee-algorithms ze-artificial intelligence, njengenjongo yokulahlekisa abantu okanye ukutshintsha intsingiselo yangaphambili. Ekwabizwa ngokuba yimidiya eveliswe yi-AI.

Itemu elihlanganisa konke lomkhiqizo owakhiwe, ukufuzisela, nokuguqulwa kwedatha nemidiya ngokuzenzakalelayo, ikakhulu ngokusebenzisa ama-algorithms wobuchwepheshe bokwakhiwe, ngenjongo yokudukisa abantu noma ukushintsha incazelo yasekuqaleni. Yaziwa nangokuthi yimidiya ekhiqizwe yi-AI.

google_research_ai_terminology_42

Transfer learning

oordragleer

Udluliselo lokufundiweyo

Ukufunda kokudlulisela

Transfer learning (TL) is a technique in machine learning (ML) in which knowledge learned from a task is re-used in order to boost performance on a related task.[1] For example, for image classification, knowledge gained while learning to recognize cars could be applied when trying to recognize trucks.

Oordragleer (OL) is ’n tegniek in masjienleer (ML) waarin kennis wat uit ’n taak geleer is, weer gebruik word om die prestasie met ’n verwante taak te verbeter.[1] Byvoorbeeld, vir prentklassifikasie kan die kennis wat verkry is terwyl daar geleer is om motors te herken, toegepas word wanneer daar geprobeer word om vragmotors te herken.

Ukufunda okudluliswayo (TL) bubuchule ekufundeni koomatshini (ML) apho ulwazi olufundwe kumsebenzi luphinda lusetyenziswe ukuze kunyuswe ukusebenza komsebenzi onxulumeneyo.[1] Umzekelo, kuhlelo lwemifanekiso, ulwazi olufunyenwe ngelixa ufunda ukuqaphela iimoto lunokusetyenziswa xa kuzanywa ukuqatshelwa iilori.

Ukufunda kokudlulisela (TL) ingubuciko ekufundeni komshini (ML) lapho ulwazi olutholwe emsebenzini lusetshenziswa kabusha ukuze kuthuthukiswe ukusebenza komsebenzi ohlobene.[1] Ngokwesibonelo, ekuhlungeni isithombe, ulwazi olutholwe ngesikhathi kufundwa ukuqaphela izimoto lungasebenza uma kuzanywa ukuqaphela amaloli.

google_research_ai_terminology_43

Unsupervised learning

leer sonder toesig

Ukufunda okungajongwangwa

Ukufunda okungagadiwe

A framework in machine learning where, in contrast to supervised learning, algorithms learn patterns exclusively from unlabeled data.

’n Raamwerk in masjienleer waar – anders as met leer onder toesig – algoritmes patrone eksklusief uit ongemerkte data leer.

Isakhelo ekufundweni koomatshini apho, ngokuchaseneyo nokufunda okugadiweyo, ii-algorithm zifunda iipateni ngokukodwa kwiidatha ezingabhalwanga.

Ifremuwekhi ekufundeni komshini lapho, ngokungafani nokufunda okugadiwe, ama-algorithms afunda amaphethini ngokukhethekile kudatha engenalebula.

google_research_ai_terminology_44

Zero-shot prompting

geenvoorbeeldinstruksieteks

Uyalelo olungena mizekelo

Ukuyalela ngeze

A prompting strategy that doesn’t include any examples about the desired input and output

’n Porboodskapstrategie wat geen voorbeelde van die verlangde invoer en uitvoer insluit nie

Isicwangiso sokuyalela esingabandakanyi nayiphi na imizekelo ngokufakwayo okufunekayo nesiphumo.

Isu lokukhumbuza elingabandakanyi noma iziphi izibonelo ezimayelana nokufakwayo kanye nomphumela ofiswayo

google_research_ai_terminology_45

A/B testing

A/B-toetsing

Uvavanyo lweA/B

Ukuhlolakuqhathanisa imikhiqizo

A statistical way of comparing two (or more) techniques—the A and the B. Typically, the A is an existing technique, and the B is a new technique. A/B testing not only determines which technique performs better but also whether the difference is statistically significant.

’n Statistiese manier om twee (of meer) tegnieke te vergelyk - die A en die B. Gewoonlik is die A ’n bestaande tegniek en die B ’n nuwe tegniek. A/B-toetse bepaal nie net watter tegniek beter presteer nie, maar ook of die verskil wel statisties relevant is.

Indlela yezibalo yokuthelekisa iindlela ezimbini (okanye ngaphezulu)-i-A kunye ne-B. Ngokuqhelekileyo, i-A yinkqubo ekhoyo, kwaye i-B yindlela entsha. Uvavanyo lwe-A/B alubonisi nje kuphela ukuba boluphi na ubugcisa obusebenza ngcono kodwa nokuba umahluko ubalulekile ngokwezibalo.

Indlela yezibalo yokuqhathanisa izindlela - u-A kanye no-B. Ngokujwayelekile, u-A uyindlela ekhona, bese u-B abe yindlela entsha. Ukuhlola kwe-A/B akugcini ngokunquma ukuthi iyiphi indlela esebenza kangcono kodwa nokuthi umehluko wezibalo ubalulekile na.

google_research_ai_terminology_46

Ablation

ablasie

Ukukhutshwa

ukukhipha izakhi

In artificial intelligence (AI), particularly machine learning (ML), ablation is the removal of a component of an AI system. An ablation study investigates the performance of an AI system by removing certain components to understand the contribution of the component to the overall system.

In kunsmatige intelligensie (KI), spesifiek masjienleer (ML), is ablasie die verwydering van ’n komponent van ’n KI-stelsel. ’n Ablasiestudie ondersoek die werkverrigting van ’n KI-stelsel deur sekere komponente te verwyder om die bydrae van die komponent tot die algehele stelsel te verstaan.

Kubukrelekrele obenziweyo (iAI), ngokukodwa ukufundwa komatshini (ML), ukukhutshwa kukukhutshwa kwecandelo lenkqubo yeAI. Uphononongo lokuchithwa luphanda ukusebenza kwenkqubo yeAI ngokususa amacandelo athile ukuqonda igalelo lecandelo kwinkqubo yonke.

Ubuchwepheshe obusebenza ngokungasikho okwemvelo (i-AI), ikakhulukazi ukufunda ngomshini (i-ML), ukukhipha izakhi zohlelo lwe-AI. Ucwaningo lwezakhi luhlola ukusebenza kohlelo lwe-AI ngokususa izakhi ezithile ukuze uqonde igalelo lezakhi ohlelweni lulonke.

google_research_ai_terminology_47

AI Adoption

KI-ingebruikneming

Uguqulelo kwiAI

Ukwamukela Ubuhlakani Obakhiwe

In computing, adoption means the transfer (conversion) between an old system and a target system in an organization (or more broadly, by anyone). In regard to AI, it means using artificial intelligence to improve workflows and decision-making processes

In rekenaarwese beteken ingebruikneming die oordrag (omskakeling) tussen ’n ou stelsel en ’n teikenstelsel in ’n organisasie (of wyer nog, deur enigiemand). Met betrekking tot KI beteken dit dat kunsmatige intelligensie gebruik word om werkvloei en besluitnemingprosesse te verbeter.

Ekusebenzeni kweekhompyutha, uguqulelo luthetha ukudluliselwa (ukuguqulwa) phakathi kwenkqubo endala nenkqubo eyithagethi kwintlangano (okanye ngokubanzi, nguye nabani na). Ngokuphathelele iAI, kuthetha ukusebenzisa ubukrelekrele obenziweyo ukuze kuphuculwe ukuhamba komsebenzi neenkqubo zokwenza izigqibo.

Kwezocwaningo lwamakhompyutha, ukuguqukela kusho ukudlulisa (ukushintsha) usuke ohlelweni oludala kanye nohlelo okuphokophelwe kulo esikhungweni (noma kabanzi, okwenziwa yinoma wubani). Maqondana ne-AI, kusho ukusebenzisa ubuchwephe obusebenza ngokungasikho okwemvelo ukuthuthukisa umsebenzi kanye nezinqubo zokuthatha izinqumo

google_research_ai_terminology_48

Anomaly detection

afwykingbespeuring

Ufunyaniso lokungaqhelekanga

Ukuhlonza okungavamile

In machine learning, this is the process of using machine learning models to identify anomalies rapidly. This serves several purposes, whether to maintain clean, high-quality data that you will use for processing or specific business purposes.

In masjienleer is dit die proses waardeur masjienleermodelle gebruik word om onreëlmatighede vinnig te identifiseer. Dit dien verskeie doele, onder andere die behoud van skoon, hoëgehaltedata wat gebruik word vir prosessering of spesifieke besigheidsdoeleindes.

Ukufunda koomatshini, le yinkqubo yokusebenzisa iimodeli zokufunda ngoomatshini ukuchonga izinto ezingaqhelekanga ngokukhawuleza. Oku kusebenzela iinjongo ezininzi, nokuba kukugcina idatha ecocekileyo, ekumgangatho ophezulu oya kuyisebenzisela ukuprosesa okanye iinjongo ezithile zoshishino.

Ekufundeni ngemishini, lokhu kuyinqubo yokusebenzisa izinhlelo zokufunda ngemishini sokuhlonza okungavamile ngokushesha. Lokhu kusebenza ngokwezinhloso ezihlukene, okungaba wukugcina imininingwane engenasici, eseqophelweni eliphezulu ozoyisebenzisela ukulungisa noma izinhloso zebhizinisi ezibaluliwe.

google_research_ai_terminology_49

Artifical general intelligence

kunsmatige algemene intelligensie

Ubukrelekrele obenziweyo ngokubanzi

Ubuhlakani Obakhiwe Obuvamile

A non-human mechanism that demonstrates a broad range of problem solving, creativity, and adaptability. For example, a program demonstrating artificial general intelligence could translate text, compose symphonies, and excel at games that have not yet been invented.

’n Niemenslike meganisme wat ’n algemene, wye reeks probleemoplossing, kreatiwiteit en aanpasbaarheid demonstreer. Byvoorbeeld, ’n program wat kunsmatige intelligensie demonstreer kan teks vertaal, simfonieë komponeer en uitblink in speletjies wat nog nie geskep is nie.

Indlela engeyoyamntu ebonisa uluhlu olubanzi lokusombulula iingxaki, ukuyila, nokuguquguquka. Umzekelo, iprogramu ebonisa ubukrelekrele obenziweyo ngokubanzi bunokuguqulela umbhalo, buqambe iisimfoni, buze bugqwese kwimidlalo engekaveliswa.

Inqubo engeziwa abantu yokuveza uhla olubanzi lokuxazulula inkinga, ubuciko, kanye nokuhambisana nezimo. Isibonelo, uhlelo olubonisa ubuchwepheshe obusebenza ngokungasikho okwemvelo jikelele obungakwazi ukuhumusha umbhalo, buqambe imiculo, futhi buhambe phambili emidlalweni engakasungulwa.

google_research_ai_terminology_50

Artificial Narrow Intelligence

kunsmatige beperkte intelligensie

Ubukrelekrele oBenziweyo oBumxinwa

Ubuhlakani Obakhiwe Obubhuncene

ANI / Narrow AI can be classified as being “limited to a single, narrowly defined task. Most modern AI systems would be classified in this category.” Some examples of narrow AI are self-driving cars and robot systems used in the medical field.

KBI / Beperkte KI kan geklassifiseer word as "beperk tot ’n enkel, noukeurig gedefinieerde taak". Meeste moderne KI-stelsels sal in hierdie kategorie geklassifiseer word". ’n Paar voorbeelde van beperkte KI is selfbestuurkarre en robotstelsels wat in die mediese veld gebruik word.

ANI / I-AI Emxinwa inokuhlelwa “njengesikelwe umda kumsebenzi omnye ochaziweyo. Uninzi lweenkqubo ze-AI zanamhlanje zingahlelwa kolu didi". Eminye imizekelo ye-AI emxinwa ziimoto eziziqhubayo neenkqubo zerobhothi ezisetyenziswa kwicandelo lezonyango.

I-AI encinyanane okungathiwa "inomkhawulo wokwenza umsebenzi owodwa umcinyane ochaziwe. Izinhlelo ze-AI zesimanje eziningi zingafakwa kulesi sigaba." Ezinye izibonelo zobuchwepheshe obusebenza ngokungasikho okwemvelo obuncinyane yizimoto ezimoto ezizishayelayo kanye nezinhlelo zamarobhothi ezisetshenziswa emkhakheni wezokwelapha.

google_research_ai_terminology_51

AutoML

OutoML

iAutoML

Ukufunda Kwemishini Ngokuzenzekela

Automated Machine Learning / Automated machine learning (AutoML) is the process of automating the tasks of applying machine learning to real-world problems. It is the combination of automation and ML.

outomatiese masjienleer / Outomatiese masjienleer (OutoML) is die proses waardeur die take waarin masjienleer op regtewêreldprobleme toegepas word, geoutomatiseer word. Dis die kombinasie van outomatisering en masjienleer.

ukuFunda ngoMatshini okuZenzekelayo / Ukufunda ngomatshini ngokuzenzekelayo (iAutoML) yinkqubo yokuzenzela imisebenzi yokufaka umatshini wokufunda kwiingxaki zehlabathi lokwenyani. Yindibaniselwano yokuzenzela neML.

Ufunda kwemishini ngokuzenzekela (i-AutoML) yinqubo yokuhlela ukuba imisebenzi izenzekele usebenzisa ukufunda kwemishini ezinkingeni zasemhlabeni zangempela. Kuhlanganisa ukuzenzekela kanye ne-ML.

google_research_ai_terminology_52

Big data

groot data

Idatha enkulu

Imininingo emikhulu

Primarily refers to data sets that are too large or complex to be dealt with by traditional data-processing application software. Big data usually includes data sets with sizes beyond the ability of commonly used software tools to capture, curate, manage, and process data within a tolerable elapsed time.

Verwys hoofsaaklik na datastelle wat te groot of kompleks is om deur tradisionele appsagteware vir dataprosessering gehanteer te word. Grootdata sluit gewoonlik datastelle in met groottes bó die vermoëns van sagtewarenutsgoed wat gewoonlik gebruik word om data binne ’n draaglike tydsduur vas te vang, saam te stel, te bestuur en te prosesseer.

Ngokuyintloko ibhekisa kwiiseti zedatha ezinkulu kakhulu okanye ezintsokothileyo ukuba zingasetyenzwa ngokwesiko lenkqubo yedatha yesoftware. Idatha enkulu idla ngokubandakanya iiseti zedatha ezinobungakanani obungaphaya kwamandla ezixhobo zesoftware ezisetyenziswa ngokuqhelekileyo ukubamba, ukulungelelanisa, ukulawula, nokucwangcisa idatha kwixesha elinyamezelekayo.

Kusho uhlu lwemininingwane olukhulu noma oludidayo oluzobhekana nesoftiwe yohlelo lokulungisa imininingwane lakudala. Imininingwane emikhulu ibandakanya uhlu lwemininingwane enobungakho obungaphezul kwamandla eamathuluzi esoftiwe evame ukusetshenziswa okuqopha, ukuphatha, ukulawula, nokulungisa imininingwane esikhathini esidlulile esamukelekile.

google_research_ai_terminology_53

Binary classification

binêre klassifisering

Ukuhlelwa kweBinary

ukuhlela kokukabili

A type of classification task that predicts one of two mutually exclusive classes: the positive class and the negative class. For example, a model that determines whether email messages are spam (the positive class) or not spam (the negative class).

’n Tipe klassifiseringtaak wat een of twee onverenigbare klassifiserings voorspel: die positiewe klas en die negatiewe klas. Byvoorbeeld, ’n model wat bepaal of ’n e-posboodskap gemorspos is (die positiewe klas) of nie gemorspos is nie (die negatiewe klas).

Uhlobo lomsebenzi wokuhlelwa oqikelela enye yeeklasi ezimbini ezizimeleyo: udidi oluhle kunye nodidi olubi. Umzekelo, imodeli emisela ukuba imiyalezo yeimeyile isispem (iklasi elungileyo) okanye ayisiso (udidi olubi).

Uhlobo lomsebenzi wokuhlukanisa ngokwamazinga oluqagela izinga elilodwa noma amabili akhethekile: amazinga okuhle kanye namazinga okubi. Isibonelo, uhlelo olunquma ukuthi ingabe imiyalelo yama-imeyli iwugaxekile (izinga lokuhle) noma akusiwo ugaxekile (izinga lokubi).

google_research_ai_terminology_54

Black Box AI

swartboks-KI

Ibhokisi yeAI eMnyama

Ubuhlakani Obakhiwe obungeqondwe abantu

A term used to describe artificial intelligence systems whose internal workings and decision-making processes are opaque or not easily understandable to humans.

’n Term wat gebruik word om KI-stelsels te beskryf waarvan die interne bewerkinge en besluitnemingprosesse ondeursigtig is en nie maklik deur mense verstaan kan word nie.

Ibinzana elisetyenziselwa ukuchaza iinkqubo zobukrelekrele obenziweyo apho iindlela zangaphakathi zokusebenza kwazo neenkqubo zazo zokwenza izigqibo zingacacanga okanye zingaqondwa lula ngabantu.

Itemu elisetshenziselwa ukuchaza ubuchwepheshe obusebenza ngokungasikho okwemvelo ezinezindlela zokusebenza zangaphakathi kanye nokuthatha izinqumo ezingacacile noma ezingaqondwa abantu kabula.

google_research_ai_terminology_55

Data minimization

dataminimalisering

unciphisozinkcukacha

Ukunciphisa imininingo

Data minimization is the principle of collecting, processing and storing only the necessary amount of personal information required for a specific purpose.

Dataminimalisering is die prinsiep waarin net die persoonlike inligting wat vir ’n spesifieke doel nodig is, ingesamel, verwerk en geberg word.

Ukunciphisa idatha ngumgaqo wokuqokelela, ukuprosesa nokugcina kuphela isixa esifunekayo senkcazelo yomntu efunekayo ngenjongo ethile.

Ukunciphisa imininingwane kuyinqubo yokuqoqa, ukulungisa nokugcina isibalo esidingekayo solwazi lomuntu siqu oludingeka ngokwenhloso ebaluliwe.

google_research_ai_terminology_56

Debiasing

vooroordeelverwydering

Ukunciphisa ukuthatha icala

Ukuchezuka kokujwayelekile

Debiasing is the reduction of bias, particularly with respect to judgment and decision making. Biased judgment and decision making is that which systematically deviates from the prescriptions of objective standards such as facts, logic, and rational behavior or prescriptive norms.

Vooroordeelverwydering is die vermindering van bevooroordeling, spesifiek met betrekking tot opinies en besluitneming. Bevooroordeelde opinies en besluitneming is dít wat stelselmatig afwyk van die voorgeskrewe objektiewe standaarde, soos feite, logikia en rasionele optrede of voorgeskrewe norme.

Ukunciphisa ukuthatha icala kukunciphisa umkhethe, ngakumbi ngokubhekisele kwisigwebo nokwenza izigqibo. Isigwebo nokwenziwa kwezigqibo okunomkhethe kukusuka ngokucwangcisiweyo kwimimiselo yemigangatho yenjongo efana neenyani, ingqiqo, nokuziphatha okunengqiqo okanye imigaqo emiselweyo.

Ukususa ukuchema kuwukunciphsa ukuchema, ikakhulukazi maqondana nokwahlulela kanye nokuthatha izinqumo. Ukwahlulela ngokungachemi kanye nokuthatha izinqumo kuyilokho okuchezuka ngokohlelo oluthile kulokho okumisiwe emigomweni yezimpokophelo njengamaqiniso, okuqondakalayo, kanye nokuziphatha ngokufanele noma izindlela ezimisiwe.

google_research_ai_terminology_57

Differential Privacy

differensiële privaatheid

Ukugcinwa kuyimfihlo kokwahlukileyo

Ubumfihlo Obuvikelayo

Differential privacy (DP) is a mathematically rigorous framework for releasing statistical information about datasets while protecting the privacy of individual data subjects. It enables a data holder to share aggregate patterns of the group while limiting information that is leaked about specific individuals.

Differensiële privaatheid (DP) is ’n wiskundige, strawwe raamwerk vir die vrystelling van statistiese inligting oor datastelle, terwyl die privaatheid van individuele data-onderwerpe beskerm word. Dit maak dat ’n datahouer aggregaatpatrone van die groep kan deel en terselfde tyd die inligting kan beperk wat deursyfer oor die spesifieke individue.

Ukugcinwa kokwahlukileyo kuyimfihlo (iDP) sisikhokelo semathematika esingqongqo sokukhupha inkcazelo yeenkcukacha manani ngeeseti zedatha ngelixa kukhusela ubumfihlo bedatha yento nganye. Kuvumela umnini wedatha ukuba abelane ngeepateni ezidityanisiweyo zeqela ngelixa enciphisa inkcazelo evuzisiweyo ngabantu abathile.

Inqubomogomo yokwahlukanisa (i-DP) iwuhlelo olunzulu lwezibalo lokukhipha ulwazi lwezibalo ngohlu lwemininingwane kube kuvikelwa imininingwane eyimfihlo yabantu abathile. Yenza umnikazi wemininingwane akwazi ukwabelana ngamaphethini ngokwezibalo ebe enciphisa ulwazi oluputshukayo ngabantu abathile.

google_research_ai_terminology_58

Edge Computing

periferale berekening

Usondezomphethelweni ngekhompuyutha

Ucubungulokusondeza

Edge computing is a distributed computing model that brings computation and data storage closer to the sources of data. Edge is about processing data closer to where it's being generated, enabling processing at greater speeds and volumes, leading to greater action-led results in real time.

Periferale berekening is ’n verspreide berekeningmodel wat berekeninge en databerging nader aan databronne bring. Periferaal behels dat data geprosesseer word nader aan waar dit gegenereer word. Dit maak dit moontlik om teen hoër spoed en groter volumes data te prosesseer, en daardeur intyds beter handelinggedrewe resultate te kry.

Ukusetyenziswa kwekhompyutha emphethelweni yimodeli yekhompyutha esasazwayo ezisa ukusetyenziswa kwekhompyutha nokugcinwa kwedatha kufutshane nemithombo yedatha. Emphethelweni umalunga nokusetyenzwa kwedatha kufutshane nalapho yenziwa khona, ivumela ukuproseswa ngezantya ezinkulu nemithamo, ekhokelela kwiziphumo ezinkulu ezikhokelwa yintshukumo ngexesha lokwenyani.

Ucubungulokuzondeza lwukudlulisela uhlelo locwaningo lwamakhompyutha olusondeza ucwaningo nokugcinwa kolwazi ndawonye nemithombo yemininingwane. Ucubungulokusondeza lumayelana nokulungisa imininingwane ibe iseduze kwalapho ikhiqizwe khona, okwenza ukulungiswa kwayo kusheshe futhi ibe miningi, okuholela emiphumeleni ebanzi ngesikhathi sangempela.

google_research_ai_terminology_59

False negative

vals negatief

Inegethivu engeyonyani

Iphutha okungelona

A result that indicates a given condition does not exist when it does. This means the algorithm generates a negative result when it should have been positive. A false negative could result in a cancerous region being overlooked, leading to a delayed diagnosis and treatment.

’n Uitslag wat aandui dat ’n gegewe kondisie nie bestaan nie terwyl dit wel bestaan. Dit beteken dat die algoritme ’n negatiewe resultaat genereer wanneer dit ’n positiewe een moes genereer. Byvoorbeeld, ’n vals negatief kan veroorsaak dat ’n kankeragtige area nie raakgesien word nie, wat kan lei tot vertraagde diagnose en behandeling.

Isiphumo esibonisa ukuba imeko echaziweyo ayikho kodwa ibe ikhona. Oku kuthetha ukuba ialgorithm ivelisa iziphumo ezinegethivu xa bekufanele ukuba ziphozithivu. Inegethivu engoyonyani inokukhokelela ekubeni ummandla onomhlaza ungahoywa, nto leyo ekhokelela kuxilongo nonyango olulibazisekileyo.

Umphumela oveza ukuthi isimo esithile asikho sibe sikhona. Lokhu kusho ukuthi i-algorithm ikhiqiza umphumela omubi kube kumele ikhiphe omuhle. Umphumelakukhohlisa ungadala ukuthi indawo enomdlavuza inganakwa, okuholela ekuphuzeni kokutholakala nokwelashwa kwesifo.

google_research_ai_terminology_60

False positive

vals positief

Iphozithivu engeyonyani

Iphutha okuyilo

A result that indicates a given condition does exist when it does not. For instance, in a medical imaging scenario, a false positive might lead to a healthy region being flagged as cancerous, potentially causing unnecessary concern and medical procedures for the patient.

’n Uitslag wat aandui dat ’n gegewe kondisie bestaan terwyl dit nie bestaan nie. Byvoorbeeld, in ’n mediesebeeldscenario kan ’n vals positief veroorsaak dat ’n gesonde area gevlag word as kankeragtig, wat onnodige bekommernis en mediese prosedures vir die pasient kan veroorsaak.

Isiphumo esibonisa ukuba imeko echaziweyo ikhona kodwa ibe ingekho. Umzekelo, kwimeko yomfanekiso wezonyango, iphozithivu ongeyonyani unokukhokelela ekubeni ummandla osempilweni kuthiwe unomhlaza, nto leyo enokubangela inkxalabo neenkqubo zonyango ezingeyomfuneko kwisigulana.

Umphumela oveza ukuthi isimo esibaluliwe sikhona sibe singekho. Isibonelo, lapho kuskenwa umzimba ukuze welashwe, umphumela wokuhle oyinkohliso ungaholela ekutheni indawo ephilile kuthiwe inomdlavuza, okungadala ukukhathazeka kanye nequbo yokwelashwa kwesidingo kungenasidingo.

google_research_ai_terminology_61

Federated Learning

gefedereerde leer

Ukufunda okudityanisiweyo

Ukuthuthukisa ubuchwepheshe

A distributed machine learning approach that trains machine learning models using decentralized examples residing on devices such as smartphones. In federated learning, a subset of devices downloads the current model from a central coordinating server. The devices use the examples stored on the devices to make improvements to the model. The devices then upload the model improvements (but not the training examples) to the coordinating server, where they are aggregated with other updates to yield an improved global model.

’n Verspreide masjienleerbenadering wat masjienleermodelle oplei met gedesentraliseerde voorbeelde wat op toestelle soos slimfone gestoor is. In gefedereerde leer laai ’n substel toestelle die huidige model van ’n sentrale, koördineringbediener af. Die toestelle gebruik die voorbeelde wat op die toestelle gestoor is om verbeteringe aan die model te maak. Die toestelle laai dan die modelverbeteringe op (maar nie die opleidingvoorbeelde nie) na die koördineringbediener toe, waar dit alles saam met ander opdaterings saamgevoeg word om ’n verbeterde algehele model te lewer.

Indlela yokufunda koomatshini esasazwayo eqeqesha iimodeli zokufunda ngoomatshini kusetyenziswa imizekelo ebekiweyo ehlala kwizixhobo ezinjengeesmartphone. Kwimfundo edibeneyo, iseti engezantsi yezixhobo ikhuphela imodeli yangoku kwiseva ephakathi yokulungelelanisa. Izixhobo zisebenzisa imizekelo egcinwe kwizixhobo ukuze kuphuculwe imodeli. Emva koko izixhobo zifaka ukuphuculwa kwemodeli (kodwa kungekhona imizekelo yoqeqesho) kwiseva yokulungelelanisa, apho zidityaniswe nolunye uhlaziyo ukuvelisa imodeli ephuculweyo yehlabathi.

Indlela yokufunda ngomshini edluliswayo eqeqesha izinhlelo zokufunda ngemishini kusetshenziswa izibonelo eziqeqelwe ndawonye kumadivayisi anjengamasamthifoni. Lapho kufundwa ngokubuchwepheshe, uhlu lwamadivayisi akhethiwe adawuniloda uhlelo olukhona ngokususela kuseva edidiyele eqoqele konke ndawonye. Amadivayisi asebenzisa izibonelo ezigcinwe kumadivayisi ukuthuthukisa uhlelo. Amadivayisi abe esefaka okuthuthukisiwe ohlelweni (kodwa hhayi isibonelo zokuqeqesha) kuseva edidiyelayo, lapho ikalwe khona nokunye ukuthuthukisa ukuze kuphume uhlelo oluthuthukisiwe lomhlaba jikelele.

google_research_ai_terminology_62

Feedback Loop

terugvoerkring

Iluphu yengxelo

Iwuku Lezimvo

Feedback occurs when outputs of a system are routed back as inputs as part of a chain of cause-and-effect that forms a circuit or loop. The system can then be said to feed back into itself. For example, if the AI system is a chatbot, it might receive feedback from users on how well it is able to understand and respond to their queries. The AI system then uses this feedback to adjust its algorithms and improve its performance.

Terugvoering vind plaas wanneer ’n stelsel se uitsette terugherlei word as insette as deel van ’n oorsaak-en-gevolgkettingverloop wat dan ’n kring of lusloop veroorsaak. Daar kan dus gesê word dat die stelsel in ditself in terug voer. Byvoorbeeld, as die KI-stelsel ’n kletsbot is, kan dit dalk terugvoer kry vannaf gebruikers oor hoe goed dit hul navrae kan verstaan en daarop reageer. Die KI-stelsel gebruik dan hierdie terugvoer om sy algoritmes aan te pas en sy werkverrigting te verbeter.

Ingxelo yenzeka xa iziphumo zenkqubo zibuyiselwa umva njengenxalenye yekhonkco lesizathu nesiphumo esenza isekethe okanye iluphu. Ngoko kunokuthiwa inkqubo ibuyela kuyo ngokwayo. Umzekelo, ukuba inkqubo ye-AI yi-chatbot, inokufumana impendulo kubasebenzisi malunga nokuba ikwazi kangakanani ukuqonda nokuphendula imibuzo yabo. Emva koko, inkqubo yeAI isebenzisa le ngxelo ukulungisa iialgorithm zayo nokuphucula ukusebenza kwayo.

Ukuthola izimvo kwenzeka lapho okufakiwe ohlelweni kubuyiselwa njengezimvo njengengxenye yochungechunge lokwenzekile nomphumela olwakha isekhethi noma iwuku. Uhlelo lungathathwa ngokuthu luyizimvo ngokwalo. Isibonelo, uma uhlelo lwe-AI luyichatbot, lungase luthola izimvo ezivela kubasebenzisi ngokuthi lukwazi kangakanani ukuqonda nokuphendula imibuzo yalo. Uhlelo lwe-AI lube selusebenzisa lezi zimvo ukuchibiyela ama-algorithm futhi kuthuthukiswe ukusebenza kwawo.

google_research_ai_terminology_63

Hyperparameter

hiperparameter

Ihyperparameter

Ilungiselelakuqeqesha

The variables that you or a hyperparameter tuning service adjust during successive runs of training a model. For example, learning rate is a hyperparameter. You could set the learning rate to 0.01 before one training session. If you determine that 0.01 is too high, you could perhaps set the learning rate to 0.003 for the next training session.

Die veranderlikes wat jy of ’n hipergrensinstellingdiens aanpas tydens agtereenvolgende opleidingsessies van ’n model. Byvoorbeeld, leerkoers is ’n hipergrens. Jy kan die leerkoers op 0.01 stel voor een opleidingsessie. As jy bepaal dat 0.01 te hoog is, kan jy die leerkoers dan dalk na 0.003 toe stel vir die volgende opleidingsessie.

Izinto eziguquguqukayo ozihlengahlengisa ngokwakho okanye yinkonzo yehyperparameter tuning ngexesha loqeqesho olulandelelanayo lwemodeli. Umzekelo, izinga lokufunda yi-hyperparameter. Unokuseta ireyithi yokufunda ibe yi-0.01 phambi kweseshoni yoqeqesho enye. Ukuba ufumanisa ukuba u-0.01 uphezulu kakhulu, usenokuseta ireyithi yokufunda ibe ngu-0.003 kwiseshoni yoqeqesho elandelayo.

Izakhi ezisetshenziswa nguwe noma ihyperparameter ngasikhathi lapho uqeqesha uhlelo nempumelelo, Isibonelo, izinga lokufunda liyihyperparameter. Ungahlela izinga lokufunda libe u-0.01 ngaphambi kwesikhathi sokufundisa esisodwa. Uma unquma ukuthi u-0.01 uphezulu kakhulu, ungase umise izinga lokufunda ku-0.003 esikhathini sokuqeqesha esilandelayo.

google_research_ai_terminology_64

Image recognition

beeldherkenning

Ukuqondwa komfanekiso

Ukubona isithombe

Also known as image classification, is a process that classifies object(s), pattern(s), or concept(s) in an image.

Ook bekend as prentklassifisering. Dit is ’n proses wat voorwerp(e), patroon(e), of konsep(te) in ’n prent klassifiseer.

Ikwaziwa njengohlelo lomfanekiso, yinkqubo ehlela into (izinto), ipatheni (iipatheni), okanye inkalo (iinkalo) kumfanekiso.

Kubuye kwaziwe ngokuthi uhlelo lokubeka izithombe ngokwamazinga, inqubo ebeka izinto, amaphethini, noma imiqondo ngokwamazinga esithonjeni.

google_research_ai_terminology_65

Intelligent agent

intelligente agent

Iarhente yobukrelekrele

Isimeleli esikhaliphile

IA / An agent that perceives its environment, takes actions autonomously in order to achieve goals, and may improve its performance with learning or acquiring knowledge. An intelligent agent may be simple or complex e.g. a thermostat, a human being...

IA / ’n Agent wat sy omgewing waarneem, voer selfstandig handelinge uit om doelwitte te bereik, en kan sy werkverrigting verbeter deur te leer of kennis in te win. ’n Intelligente agent kan eenvoudig of kompleks wees, byvoorbeeld ’n termostaat, ’n mens …

IA / Iarhente ebona okuyingqongileyo, ithatha amanyathelo ngokuzimeleyo ukuze ifezekise iinjongo, kwaye inokuphucula ukusebenza kwayo ngokufunda okanye ngokufumana ulwazi. Iarhente ekrelekrele isenokuba lula okanye intsonkothe umz. ithermostat, umntu..

Isimeleli esiqonda indawo esikuyo, esithatha isinyathlo ngokuzenzekela ukuze kuveze imogomo, futhi kuthuthukiswe ukusebenza ngokufunda noma ukuthola ulwazi. Isimeleli esikhaliphile singaba lula noma nzulu isib. isiklalizingakushisa, umuntu

google_research_ai_terminology_66

Memory leak

geheuelek

Ukuvuza kwememori

Ukuputshuka kolwazi olugciniwe

In computer science, a memory leak is a type of resource leak that occurs when a computer program incorrectly manages memory allocations in a way that memory which is no longer needed is not released. It reduces the performance of the computer by reducing the amount of available memory.

In rekenaarwetenskap is geheuesyfering ’n tipe hulpbronverlies wat voorkom wanner ’n rekenaarprogram verkeerdelik geheuetoewysing bestuur op só ’n wyse dat geheue wat nie meer benodig word nie, nie vrygestel word nie. Dit verminder die werkverrigting van die rekenaar deur die beskikbare geheue te verminder.

Kwisayensi yekhompyutha, ukuvuza kwememori luhlobo lokuvuza kwezixhobo okwenzekayo xa inkqubo yekhompyutha ilawula ngokungalunganga ukwabiwa kweememori ngendlela yokuba imemori engasadingekiyo ingakhutshwa. Kunciphisa ukusebenza kwekhompyutha ngokunciphisa inani lememori ekhoyo.

Kwezesayensi yamakhompyutha, ukuputshuka kolwazi kuwuhlobo lokuputshuka kwemithombo okwenzeka lapho uhlelo lwekhompyutha lulawula ukwabiwa kolwazi olugciniwe ngokungafanele ngendlela yokuthi ulwazi olugciniwe olungasadingeki lungakhishwa. Lunciphisa ukusebenza kwekhompyutha ngokunciphisa ubungako bolwazi olugciniwe olukhona.

google_research_ai_terminology_67

Overfitting

oormatige passing

Ukungalingani okugqithileyo

Ukuhambisana ngqo

In mathematical modeling, overfitting is "the production of an analysis that corresponds too closely or exactly to a particular set of data, and may therefore fail to fit to additional data or predict future observations reliably"

In wiskundige modelleerwerk is oorpassing “die produsering van ’n ontleding wat te nou of presies ooreenstem met ’n datastel en daarom nie betroubaar by bykomende data kan pas of toekomstige waarnemens kan voorspel nie.

Kwimodeli yemathematika, ukulingana ngokugqithisileyo "kukuveliswa kohlalutyo oluhambelana ngokusondeleyo okanye ngokuthe ngqo kwiseti ethile yedatha, kwaye ke ngoko inokusilela ukulingana nedatha eyongezelelweyo okanye ukuqikelela imigqaliselo yexesha elizayo ngokuthembekileyo"

Ohlelweni lwezibalo, ukuhambisana ngqo "kuwukukhiqizwa kocwaningo oluhambisana kakhulu noma ngqo nohlu lweminingwane ethile, futhi olungase lwahluleke ukungena kweminye imininingwane noma ukuqagela okuzohlolwa ngokuzayo ngokwethembeka"

google_research_ai_terminology_68

Oversampling

opbalansering

Iisampuli egqithisileyo

Ukusetshenziswa kwamasampula kabusha

Reusing the examples of a minority class in a class-imbalanced dataset in order to create a more balanced training set.

Die hergebruik van voorbeelde van ’n minderheidklas in ’n klasongebalanseerde datastel om sodoende ’n meer gebalanseerde opleidingstel te skep.

Ukusebenzisa kwakhona imizekelo yeklasi encinci kwiseti yedatha kwiklasi yokungalingani ukwenzela ukudala isethi yoqeqesho olulungeleleneyo.

Ukusebenzisa kabusha izibonelo ezingeni lokuncane ohlwini lwemininingwane olungalingene ngokufanele ukuze kwakhiwe uhlu lokuqeqesha olulingene.

google_research_ai_terminology_69

Precision

presisie

Ukuchana

ukuqonda ngqo

Precision is the fraction of relevant instances among the retrieved instances. It is calculated by dividing relevant retrieved instances with all retrieved instances. For example, a computer program for recognizing dogs (the relevant element) in a digital photograph processes a picture which contains 10 cats and 12 dogs. The program then identifies 8 dogs. Of the eight elements identified as dogs, only 5 actually are dogs (true positives), while the other 3 are cats (false positives). 7 dogs were missed (false negatives), and 7 cats were correctly excluded (true negatives). The program's precision is then 5/8 (true positives / selected elements).

Presisie is die breukdeel van relevante gevalle uit die gerapporteerde gevalle. Dit word uitgewerk deur die relevante gerapporteerde gevalle te deel deur al die gerapporteerde gevalle. Byvoorbeeld, ’n rekenaarprogram wat honde (die relevante element) in digitale foto's identifiseer, prosesseer ’n foto wat 10 katte en 12 honde in het. Die program identifiseer dan 8 honde. Uit die 8 elemente wat as honde geïdentifiseer is, is net 5 regtig honde (werklike positief) , terwyl 3 katte is (vals positief). 7 honde is misgekyk (vals negatief) en 7 katte is korrek uitgesluit (werklike negatief). Die program se presisie is dan 5/8 (werklike positiewe / gerapporteerde elemente).

Ukuchaneka liqhezu leemeko ezifanelekileyo phakathi kweemeko ezifunyenweyo. Ibalwa ngokwahlula iimeko ezifanelekileyo kwezifunyenweyo nazo zonke iimeko ezifunyenweyo. Umzekelo, inkqubo yekhompyutha yokuqaphela izinja (into efanelekileyo) kwifoto yedijithali iprosesa umfanekiso oqulethe iikati ezili-10 nezinja eziyi-12. Inkqubo ke iye ichonge izinja ezisi-8. Kwizinto ezisibhozo ezichongwe njengezinja, zi-5 kuphela ezizinja (iiphozithivu zokwenyani), ngelixa ezinye ezi-3 izikati (iiphozithivu ezingeyonyani). Izinja ezisi-7 ziphosiwe (iinegethivu ezingeyonyani), zaze iikati ezisi-7 zangabandakanywa ngokuchanekileyo (iinegethivu eziyinyani). Ukuchaneka kwenkqubo ngu-5/8 (iiphozithivu eziyinyani / izinto ezikhethiweyo).

Ukugqonda ngqo kuyingcosana yezimo ezifanele phakathi kwezimo ezilandiwe. Kubalwa ngokuhlukanisa izimo ezilandiwe ezifanele nazo zonke izimo ezilandiwe. Isibonelo, uhlelo lwekhompyutha lokubona izinja (isakhi esifanele) esithombeni sedijithali izinqubo zesithombe eziqukethe amakati ayi-10 nezinja eziyi-12. Uhlelo lube seluhlonza izinja eziyi-8. Ezakhini eziyisishiyagalombili ezihlonzwe njengezinja, zi-5 kuphela eziyizinja (umphumela wokuhle oyiqiniso), kube ezi-3 zingamakati (umphumela wokhlei oyinkohliso) Izinja eziyisi-7 azibonwanga (imiphumela yokubi eyinkohliso), kanye namakati ayisi-7abandakanywe ngokufanele (umphumela wokubi oyiqiniso) Ukuqonda ngqo kohlelo lube seluba ngu-5/8 (imiphumela emihle ngokweqiniso / izakhi ezikhethiwe).

google_research_ai_terminology_70

Predictive analytics

voorspelontleding

Uhlalutyo oluqikelelayo

Ucubungulo oluqagelayo

Predictive analytics is a form of analytics that uses machine learning or technology to predict what will happen in a specific time frame based on historical data and patterns.

Voorspelontleding is ’n vorm van ontleding wat masjienleer of tegnologie gebruik om volgens historiese data en patrone te voorspel wat in ’n spesifieke tydperk sal gebeur.

Uhlalutyo lokuqikelela luhlobo lohlalutyo olusebenzisa ukufundwa koomatshini okanye iteknoloji ukuze kuqikelelwe oko kuza kwenzeka kwixesha elithile ngokusekelwe kwidatha yembali neepatheni.

Ucubungulo oluqagelayo wuhlobo lwezibalo olusebenzisa ukufunda ngemishini noma ubuchwepheshe ukuqagela okuzokwenzeka esikhathi esithile esimisiwe kususelwa kumininingwane esemlandweni kanye namaphethini.

google_research_ai_terminology_71

Privacy Impact Assessment

privaatheidimpakassessering

UVavanyo lweMpembelelo yoBumfihlo

Ukuhlola Umthelela Wobumfihlo

PIA / A Privacy Impact Assessment (PIA) is a process which assists organizations in identifying and managing the privacy risks arising from new projects, initiatives, systems, processes, strategies, policies, business relationships etc. / <A Privacy Impact Assessment is a process which assists organizations in identifying and managing the privacy risks arising from new projects, initiatives, systems, processes, strategies, policies, business relationships etc.>

PIA / ’n Privaatheidimpakassessering (PIA) is ’n proses wat organisasies help om die privaatheidrisiko's te identifiseer en te bestuur wat kan voortspruit uit nuwe projekte, innisiatiewe, stelsels, prosesse, strategieë, beleide, besigheidsverhoudings ens. / <’n Privaatheidimpakassessering is ’n proses wat organisasies help om die privaatheidrisiko's te identifiseer en te bestuur wat kan voortspruit uit nuwe projekte, innisiatiewe, stelsels, prosesse, strategieë, beleide, besigheidsverhoudings ens.>

iPIA / Uvavanyo Lwempembelelo Yobumfihlo (PIA) yinkqubo enceda imibutho ekuchongeni nasekulawuleni imingcipheko yobumfihlo evela kwiiprojekthi ezintsha, izinto ezintsha, iisistim, iinkqubo, izicwangciso, iipolisi, ubudlelwane boshishino njl.njl.

Ukuhlola Umthelela Wobumfihlo (i-PAIA) kuyinqubo esiza izinhlangano ekuhlonzeni nasekuphatheni ingcuphe yobumfihlo evela kumaphrojekthi amasha, izinhlelo kanye nezinqubo, amaqhingasu, izinqubomgomo, ubudlelwano bebhizinisi njl.

google_research_ai_terminology_72

Prompt injection

instruksieteksinvoeging

Utofomyalelo

Ukufaka imiyalelo

An attack that manipulates a large language model (LLM) by injecting malicious inputs designed to alter the model’s output. This type of cyber-attack exploits the way LLMs process and generate text based on input prompts. By inserting carefully crafted text, an attacker can trick the model into producing unauthorized content, accessing restricted data, or executing specific actions.

’n Aanval wat ’n groottaalmodel (GTM) manipuleer deur kwaadwillige invoere in te voeg wat ontwerp is om die model se uitset te verander. Hierdie soort kuberaanval byt die manier uit waarop GTM'e teks met invoerporboodskappe prosesseer en genereer. Deur fyn saamgestelde teks in te voeg, kan ’n aanvaller die model flous om dit ongoedgekeurde inhoud te laat produseer, toegang te kry tot beperkte data, of om spesifieke handelinge uit te voer.

Uhlaselo olusebenzisa imodeli yolwimi olukhulu (LLM) ngokutofa igalelo elibi eliyilelwe ukuguqula imveliso yemodeli. Olu hlobo lohlaselo lwecyber lusebenzisa indlela iiLLM eziprosesa zize zivelise ngayo itekisi ngosekelwe kumyalelo ofakiweyo. Ngokufaka itekisi eyenziwe ngobuchule, umhlaseli unokukhohlisa imodeli ukuba ivelise ikhontenti engagunyaziswanga, ifikelele kwidatha ethintelweyo, okanye ukwenza izenzo ezithile.

Ukuhlasela okusebenzisa izinhlelo zolimi ezinkulu (ama-LLM) ngokufaka okumoshayo okwakhele ukuphazamisa umphumela wohlelo. Lolu hlobo lokuhlasela nge-inthanethi kuxhaphaza indlela ama-LLM alungisa futhi akhiqize ngayo izibizelo zokufakwayo. Ngokufaka ngokucophelela umbhalo obunjiwe, umhlaseli angadida uhlelo ukuze lukhiqize ulwazi olungagunyazwanga, lufinyelele emininingwaneni evinjelwe, noma lusebenise imiyalelo emisiwe.

google_research_ai_terminology_73

Pseudonymization

pseudonimisering

IPseudonymization

Ukuzivezambumbulu

The process of obscuring data with the ability to re-identify it later. Personally identifiable information fields within a data record are replaced by one or more artificial identifiers, or pseudonyms.

Die proses om data te verdoesel met die vermoë om dit later te heridentifiseer. Velde met persoonlik identifiseerbare inligting in ’n datarekord word vervang met een of meer kunsmatige identifiseerders, of pseudonieme.

Inkqubo yokufihla idatha nokwazi ukuyichonga kwakhona kamva. Indawo yenkcazelo echongiweyo ngokobuqu kwirekhodi yedatha ethathelwe indawo sisifanisi esinye, okanye ngapezulu, okanye ii-pseudonym.

Inqubo yokufihla idatha ngendlela yokukwazi ukuyibona futhi kamuva. Izinkundla zolwazi olukhomba umuntu siqu (PII) erekhodini ledatha ziyakhishwa kufakwe izinkomba zokwenziwa ezizodwa noma ngaphezulu, noma amapseudonyms.

google_research_ai_terminology_74

Recall

volledigheid

Ukukhumbula

Isibuyiseli

Recall is the fraction of relevant instances that were retrieved. It is calculated by dividing relevant retrieved instances with all relevant instances. For example, a computer program for recognizing dogs (the relevant element) in a digital photograph processes a picture which contains 10 cats and 12 dogs. The program then identifies 8 dogs. Of the eight elements identified as dogs, only 5 actually are dogs (true positives), while the other 3 are cats (false positives). 7 dogs were missed (false negatives), and 7 cats were correctly excluded (true negatives). The program's recall is 5/12 (true positives / relevant elements).

Volledigheid is die breukdeel van relevante gevalle wat gerapporteer is. Dit word uitgewerk deur die relevante gerapporteerde gevalle te deel deur die totale aantal relevante gevalle. Byvoorbeeld, ’n rekenaarprogram wat honde (die relevante element) in digitale foto's identifiseer, prosesseer ’n foto wat 10 katte en 12 honde in het. Die program identifiseer dan 8 honde. Uit die 8 elemente wat as honde geïdentifiseer is, is net 5 regtig honde (werklike positief) , terwyl 3 katte is (vals positief). 7 honde is misgekyk (vals negatief) en 7 katte is korrek uitgesluit (werklike negatief). Die program se volledigheid is dan 5/12 (werklike positiewe / relevante elemente).

Ukukhumbula liqhezu leemeko ezifanelekileyo eziye zafunyanwa. Kubalwa ngokwahlula iimeko ezifanelekileyo kwezifunyenweyo nazo zonke iimeko ezifanelekileyo. Umzekelo, inkqubo yekhompyutha yokuqaphela izinja (into efanelekileyo) kwifoto yedijithali iprosesa umfanekiso oqulethe iikati ezili-10 nezinja eziyi-12. Inkqubo ke iye ichonge izinja ezisi-8. Kwizinto ezisibhozo ezichongwe njengezinja, zi-5 kuphela ezizinja (iiphozithivu zokwenyani), ngelixa ezinye ezi-3 izikati (iiphozithivu ezingeyonyani). Izinja ezisi-7 ziphosiwe (iinegethivu ezingeyonyani), zaze iikati ezisi-7 zangabandakanywa ngokuchanekileyo (iinegethivu eziyinyani). Ukukhumbula kwenkqubo ngu-5/12 (iiphozithivu eziyinyani / izinto ezifanelekileyo).

Ukubuyisa kuyingcosana yezimo ezifanele ezilandiwe. Kubalwa ngokuhlukanisa izimo ezilandiwe ezifanele nazo zonke izimo ezifanele. Isibonelo, uhlelo lwekhompyutha lokubona izinja (isakhi esifanele) esithombeni sedijithali izinqubo zesithombe eziqukethe amakati ayi-10 nezinja eziyi-12. Uhlelo lube seluhlonza izinja eziyi-8. Ezakhini eziyisishiyagalombili ezihlonzwe njengezinja, zi-5 kuphela eziyizinja (umphumela wokuhle oyiqiniso), kube ezi-3 zingamakati (umphumela wokhlei oyinkohliso) Izinja eziyisi-7 azibonwanga (imiphumela yokubi eyinkohliso), kanye namakati ayisi-7abandakanywe ngokufanele (umphumela wokubi oyiqiniso) Ukukhumbula kohlelo kube ngu-5/12 (imiphumela emihle ngokweqiniso / izakhi ezifanele).

google_research_ai_terminology_75

Reinforcement Learning from human feedback

versterkingleer uit menslike terugvoer

Ukufunda okukhuthazwa yingxelo yabantu

Ukufunda Ukuthatha Izinqumo ngebuyambiko elivela kubantu

In machine learning, reinforcement learning from human feedback (RLHF) is a technique to align an intelligent agent with human preferences. It involves training a reward model to represent preferences, which can then be used to train other models through reinforcement learning.

In masjienleer is verstekleer uit menseterugvoer (VLMT) ’n tegniek om ’n intelligente agent te belyn met mensevoorkeure.

Ekufundweni koomatshini, ukufunda okukhuthazwa yingxelo esuka kubantu (iRLHF) bubuchule bokulungelelanisa iarhente ekrelekrele nezinto ezikhethwa ngabantu. Ibandakanya ukuqeqesha imodeli yokuvuza ukuze imele izinto ezikhethwayo, ezinokuthi zisetyenziselwe ukuqeqesha ezinye iimodeli ngokufunda okukhuthazayo.

Ekufundeni kwemishini, ukufunda kokugcizelela okususelwa ezimvweni zabantu (i-RLHF) kuyindlela yokuqondanisa isimeleli esikhaliphile nalokho okukhethwa abantu. Kubandakanya ukuqeqesha uhlelo lwemiphumela ukuze luveze okukhethwayo, okungasetshenziselwa ukuqeqesha ezinye izindlela zokugcizelela ukufunda.

google_research_ai_terminology_76

Retrieval augmented generation

ophaalaangevulde generering

Ukuvelisa ukufumana okwandisiweyo

ukukhiqiza kokulanda ulwazi

Retrieval augmented generation (RAG) is a type of generative artificial intelligence that has information retrieval capabilities. It modifies interactions with a large language model (LLM) so that the model responds to user queries with reference to a specified set of documents, using this information in preference to information drawn from its own vast, static training data. This allows LLMs to use domain-specific and/or updated information.

Aanvullendeophaalgenerering (AOG) is ’n tipe generatiewe kunsmatige intelligensie wat inligtingophaalvermoëns het. Dit verander interaksies met ’n groottaalmodel (GTM) sodat die model reageer op navrae met verwysing na ’n gespesifiseerde stel dokumente en hierdie inligting met voorkeur gebruik in plaas van inligting wat dit vanuit sy eie wye, onveranderde opleidingdata trek. Dit maak dat die GTM domeinspesifieke en/of opgedateerde inligting kan gebruik.

Ukuvelisa ukufumana okwandisiweyo (iRAG) luhlobo lobukrelekrele obenziweyo obuvelisayo obunesakhono sokufumana inkcazelo. Ilungisa unxibelelwano nemodeli yolwimi olukhulu (LLM) ukuze imodeli iphendule imibuzo yabasebenzisi ngokubhekisele kuluhlu oluthile lwamaxwebhu, isebenzisa le nkcazelo ngokukhetha inkcazelo ethathwe kwidatha yayo enkulu, engatshintshiyo yoqeqesho. Oku kuvumela iiLLM ukuba zisebenzise idomeyini ekhethekileyo kunye/okanye inkcazelo ehlaziyiweyo.

Ukukhiqiza kokulnadna ulwazi (i-RAG) kuwuhlobo lobuchwepheshe obusebenza ngokungasikho okwemvelo obukhiqizayo olukwazi ukulanda ulwazi. Lulungisa ukuxhumana kwezinhlelo zolimi ezinkulu (ama-LLM) ukuze izinhlelo ziphendule imibuzo yomsebenzisi zisusela ulwazi ohlwini lwemibhalo, zisebenzisa lolu lwazi kunokukhetha ulwazi oluthathwe emininingwaneni yalo yokuqeqesha ebanzi ngemile. Lokhu kwenza ukuthi ama-LLM akwazi ukusebnezisa ulwazi oluqondene nomkhakha othile kanye/noma olubuyekeziwe.

google_research_ai_terminology_77

Robustness

robuustheid

Ukomelela

Ubuqatha

In computer science, robustness is the ability of a computer system to cope with errors during execution and cope with erroneous input. It is tested using techniques such as fuzz testing (an automated software testing technique that involves providing invalid, unexpected, or random data as inputs to a computer program. The program is then monitored for exceptions such as crashes or potential memory leaks.)

In rekenaarwetenskap is robuustheid die vermoë van ’n rekenaarstelsel om foute tydens uitvoering en foutiewe invoer te kan hanteer. Dit word getoets met tegnieke soos donstoetse (’n outomatiese sagtewaretoetstegniek waarin ongeldige, onverwagte of lukrake data as invoere vir ’n rekenaarprogram verskaf word. Die program word dan gemonitor vir uitsonderings soos omvalle of moontlike geheuesyferings).

Kwisayensi yekhompyutha, ukomelela kukukwazi kwenkqubo yekhompyutha ukumelana neengxaki ngexesha lokwenziwa nokujongana negalelo elinengxaki. Ivavanywa kusetyenziswa ubuchule obufana novavanyo lokufana (ubuchule bokuvavanya isoftwe ezenzekelayo ebandakanya ukunika idatha engasebenziyo, engalindelekanga, okanye engacwangciswanga njengegalelo kwinkqubo yekhompyutha. Inkqubo iye ibekwe iliso kwizinto ezisecaleni ezifana nokukhresha okanye ukuvuza kwememori okunokubakho.)

Kwezesayensi yekhompyutha, ukubekeza kungamandla ohlelo lwekhompyutha lokubekezelela amaphutha lapho lusetshenziswa futhi lubekezelele okufakwe ngephutha. Luhlolwa kusetshenziswa izindlela ezinjengokuhlola ngokungaphelele (indlela yokuhlola ngesoftiwe ngokuzenzekela okubandakanya ukuhlinzeka ngemininingwane engasebenzi, engalindelekile, noma engavamile njengokufakwa ohlelweni lwekhompyutha. Uhlelo lube seluqashwa kubhekwe okuphuma eceleni njengokushayisa noma ukuputshuka kolwazi okungase kube khona.)

google_research_ai_terminology_78

Self-training

selfopleiding

Ukuziqeqeshela

Ukuziqeqesha

A variant of self-supervised learning that is particularly useful when all of the following conditions are true: - The ratio of unlabeled examples to labeled examples in the dataset is high. - This is a classification problem.

’n Variant van self-gekontrolleerde leer wat besonders nuttig is wanneer al die volgende toestannde waar is: - Die verhouding van ongeëtiketteerde voorbeelde teenoor geëtiketteerde voorbeelde in die datastel hoog is. - Dit ’n klassifiseringprobleem is.

Umahluko wokufunda okugadwe ngokuphakathi uluncedo ngakumbi xa zonke ezi meko zilandelayo ziyinyani: - Umlinganiselo wemizekelo engabhalwanga kwimizekelo enelebhile kuluhlu lwedatha uphezulu. - Le yingxaki yokuhlelwa.

Igatsha lokuzifundela oluwusiza lapho okulandelayo kuyiqiniso: - Lokhu kuyinkinga yokubeka ngokwamazinga.

google_research_ai_terminology_79

Superintelligence

superintelligensie

Ubukrelekrele obuphezulu

Ubuhlakani obukhulukazi

A superintelligence is a hypothetical agent that possesses intelligence surpassing that of the brightest and most gifted human minds.

’n Superintelligensie is ’n hipotetiese agent wat intelligensievermoëns het wat ver bo dié van die slimste en mees begaafde menslike denke is.

Ubukrelekrele obuphezulu yiarhente ecingelwayo enobukrelekrele obudlula ezona ngqondo ziqaqambileyo nezinesiphiwo zomntu.

Ubuhlakani obukhulukazi buyisimeleli esingasiso esangempela esiqukethe ubuhlakani obudlula abantu abahlakaniphe kakhulu kakhulu.

google_research_ai_terminology_80

Temperature

temperatuur

Ubushushu

Iqaphasimo

A hyperparameter that controls the degree of randomness of a model's output. Higher temperatures result in more random output, while lower temperatures result in less random output. Choosing the best temperature depends on the specific application and the preferred properties of the model's output. For example, you would probably raise the temperature when creating an application that generates creative output. Conversely, you would probably lower the temperature when building a model that classifies images or text in order to improve the model's accuracy and consistency.

’n Hipergrens wat die mate van lukraakheid van ’n model se uitset beheer. Hoër temperature veroorsaak meer lukrake uitsette, terwyl laer temperature minder lukrake uitsette veroorsaak. Om die beste temperatuur te kies hang af van die spesifieke toepassing en die voorkeureienskappe van die model se uitset. Byvoorbeeld, jy sal waarskynlik die temperatuur verhoog wanneer jy ’n toepassing skep wat kreatiewe uitsette genereer. Omgekeerd, sal jy waarskynlik die temperatuur verlaag wanneer jy ’n model bou wat prente of teks klassifiseer om sodoende die model se akkuraatheid en konsekwentheid te verbeter.

Ihyperparameter elawula iqondo lokungakhethi kwemveliso yemodeli. Amaqondo obushushu aphezulu abangela imveliso engacwangciswanga ngakumbi, ngelixa amaqondo obushushu asezantsi ephumela kwimveliso engacwangciswanga ngokuncinci. Ukukhetha obona bushushu kuxhomekeke kwiaplikheyshini ethile neempawu ezikhethiweyo zemveliso yemodeli. Umzekelo, unganyusa ubushushu xa usenza iaplikheyshini evelisa imveliso yokuyila. Ngokuchaseneyo, ubuya kuthoba iqondo lobushushu xa usakha imodeli ehlela imifanekiso okanye umbhalo ukuze kuphuculwe ukuchaneka kwemodeli nokuhambelana.

Uhlelo lokulawula ukufunda ngemishini olulawula izinga lokungavami kwemiphumela yohlelo. Umphumela wezinga eliphezulu luwumphumela ongavamile kakhulu, kube umphumela wamadszingakushisa aphansi wumphumela ongavamile kancane. Ukukhetha izingakushisa eliphuma phambili kuncike ohlweni olubaluliwe kanye nezakhi ezikhethwayo emphumeleni wohlelo. isibonelo, ungase unyuse izingakushisa uma wakha uhlelo olukhiqiza umphumela wobuciko. Kungenjalo, ungase wehlise izingakushisa lapho wakha uhlelo olubeka izithombe ngokwamazinga ukuze kuthuthukiswe ukunemba nokuhambisana kohlelo.

google_research_ai_terminology_81

True negative

ware negatief

Inegethivu eyinyani

Okungekuhle ngokweqiniso

TN / An example in which the model correctly predicts the negative class. For example, the model infers that a particular email message is not spam, and that email message really is not spam.

WN / ’n Voorbeeld waarin die model die negatiewe klas korrek voorspel. Byvoorbeeld, die model lei af dat ’n spesifieke e-posboodskap nie gemorspos is nie, en die e-pos dan werklik nie gemorspos is nie.

TN / Umzekelo apho imodeli iqikelela ngokuchanekileyo udidi olunegethivu. Umzekelo, imodeli ichaza ukuba umyalezo othile weimeyile ayisospem, ube loo myalezo weimeyile ungesospem nyhani.

i-TN / Isibonelo sokuqagela kohlelo ngokufanele ezingeni lokungekuhle. Isibonelo, uhlelo lusho ukuthi umyalezo we-imeyli othile awusiwo ugaxekile, futhi umyalezo we-imeyli awusiwo ngempela ugaxekile.

google_research_ai_terminology_82

True positive

ware positief

Iphozithivu eyinyani

Okuhle ngokweqiniso

TP / An example in which the model correctly predicts the positive class. For example, the model infers that a particular email message is spam, and that email message really is spam.

WP / ’n Voorbeeld waarin die model die positiewe klas korrek voorspel. Byvoorbeeld, die model lei af dat ’n spesifieke e-posboodskap gemorspos is, en die e-pos dan werklik gemorspos is.

TP / Umzekelo apho imodeli iqikelela ngokuchanekileyo udidi oluphozithivu. Umzekelo, imodeli ithi umyalezo othile weimeyile sispem, ibe loo myalezo weimeyile ube usisipem nyhani.

i-TP / Isibonelo sokuqagela kohlelo ngokufanele ezingeni lokuhle. Isibonelo, uhlelo lusho ukuthi umyalezo we-imeyli othile uwugaxekile, futhi umyalezo we-imeyli ngempela uwugaxekile.

google_research_ai_terminology_83

Underfitting

onderpassing

Ukungalingani okunganeno

uhlelo olubuthaka

Producing a model with poor predictive ability because the model hasn't fully captured the complexity of the training data.

Om ’n model te produseer met swak voorspellingvermoë omdat die model nie die kompleksiteit van die opleidingdata ten volle aangeteken het nie.

Ukuvelisa imodeli enesakhono esilambathayo sokuqikelela ngenxa yokuba imodeli khange ibambe ngokupheleleyo ubunzima bedatha yoqeqesho.

Ukukhiqiza uhlelo olungaqinile kwezokuqagela ngoba uhlelo longakayiqophi imininingwane yokuqeqesha enzulu.

google_research_ai_terminology_84

Undersampling

afbalansering

Isampuli enganeno

Ukususa amasampula amakhulu

Removing examples from the majority class in a class-imbalanced dataset in order to create a more balanced training set.

Die verwydering van voorbeelde van die meerderheidklas in ’n klasongebalanseerde datastel om sodoende ’n meer gebalanseerde opleidingstel te skep.

Ukususa imizekelo kwiklasi yesininzi kwiseti yedatha kwiklasi elinganayo ukwenzela ukudala isethi yoqeqesho olulungeleleneyo.

Ukususa izibonelo ezingeni lokukhulu ohlwini lwemininingwane olungalingene ngokufanele ukuze kwakhiwe uhlu lokuqeqesha olulingene.

Datasets:

dsfsi
/

za-mafoko-ai-glossary

AI Terminologies in African Languages

Dataset Description

Dataset Structure

Files Overview

Data Format

Languages Covered

Usage

Example Applications

Data Quality and Provenance

Citation

Dataset Citation

Paper Citation

Authors and Contributors

Original Authors

Dataset Curation

Project Information

License

Contact

Acknowledgments

Collection including dsfsi/za-mafoko-ai-glossary

za-mafoko