Ukuhambisana (Alignment)
Alignment
Inqubo yokuqinisekisa ukuthi izinjongo zesistimu ye-AI, imiphumela, kanye nokuziphatha kuhambisana nezinjongo nezindinganiso zabantu. Lokhu kubaluleke kakhulu ezinhlelweni ezithuthukile ezingathuthukisa ukuziphatha okungahlosiwe ngokusobala.
Isibonelo: Ukuqinisekisa ukuthi i-chatbot yokusekela impilo yengqondo ayikaze incoma izenzo ezilimazayo kungakhathaliseki ukuthi yiziphi izikhuthazo.
Isixhumanisi Sokuhlela Uhlelo (API) (Application Programming Interface (API))
Application Programming Interface (API)
Iqoqo lemithetho echaziwe kanye nezivumelwano ezivumela izinhlelo zesofthiwe ezahlukene ukuba zixhumane futhi zishintshisane ngedatha.
Isibonelo: Ukusebenzisa i-OpenAI API ukuthumela isikhuthazo nokuthola impendulo ekhiqizwe imodeli yolimi kuhlelo lwakho lwewebhu.
Ubuhlakani Bokwenziwa Obujwayelekile (AGI) (Artificial General Intelligence (AGI))
Artificial General Intelligence (AGI)
Uhlobo oluyimfundiso lwe-AI olungenza noma yimuphi umsebenzi wobuhlakani ongawenza umuntu. Lujwayele ukufunda kuzo zonke izizinda.
Isibonelo: Uhlelo lwe-AGI lungafunda ukwakha umculo, lwenza ukuhlinzwa, futhi luphumelele ukuhlolwa kwefilosofi ngaphandle kohlelo oluthile lomsebenzi.
Ubuhlakani Bokwenziwa (AI) (Artificial Intelligence (AI))
Artificial Intelligence (AI)
Ukulingisa ubuhlakani bomuntu emishinini ehlelwe ukuba icabange, icabange, futhi yenze ngokuzimela.
Isibonelo: I-AI inika amandla abasizi bomuntu siqu abanjengoSiri nezinhlelo zokushayela ezizimele njengeTesla Autopilot.
Izimiso Zokuziphatha ze-AI (AI Ethics)
AI Ethics
Isiyalo esikhathazekile ngemiphumela yokuziphatha yokuthuthukiswa nokusetshenziswa kwe-AI, okuhlanganisa ubulungiswa, ubumfihlo, ukuziphendulela, kanye nokungabandlululi.
Isibonelo: Ukwakha imihlahlandlela yokuvimbela ama-algorithms okuqasha ukuba angabandlululi ngokobulili noma ubuhlanga.
Ubuhlakani Obukhulisiwe (Augmented Intelligence)
Augmented Intelligence
Imodeli yokubambisana lapho i-AI ihambisana futhi ithuthukise ubuhlakani bomuntu esikhundleni sokubufaka esikhundleni.
Isibonelo: Amathuluzi e-radiology anikwe amandla yi-AI agqamisa okungajwayelekile kodokotela, abenza ukuxilongwa kokugcina.
Umenzeli Ozimele (Autonomous Agent)
Autonomous Agent
Uhlelo lwe-AI olukwazi ukwenza izinqumo zalo futhi lwenze izenzo ukuze lufeze izinjongo zalo ngaphandle kokungenelela komuntu.
Isibonelo: Irobhothi lokulethwa elizishayelayo elizulazula emigwaqweni yedolobha futhi ligwema izithiyo ngokuzimela.
Ukubuyisela Emuva (Backpropagation)
Backpropagation
Indlela yokuqeqesha amanethiwekhi e-neural ngokuvuselela izisindo ngemuva kusuka kumalayeri okuphumayo kuya kumalayeri okufaka, kunciphisa amaphutha okubikezela.
Isibonelo: Isetshenziswa ekuqeqesheni izihlukanisi zezithombe ukuze kuncishiswe izinga lamaphutha ekuqashelweni kwezinombolo ezibhalwe ngesandla.
Ukukhetha (Ukukhetha Kwe-Algorithmic) (Bias (Algorithmic Bias))
Bias (Algorithmic Bias)
Ukukhetha okungahlosiwe nokuhlelekile emiphumeleni ye-AI ngenxa yedatha yokuqeqesha engalingani noma engameleli.
Isibonelo: Uhlelo lokuqashelwa kobuso olungaqasheli kahle abantu abamnyama kaningi ngenxa yokungameleleki okuncane kwedatha yokuqeqesha.
Idatha Enkulu (Big Data)
Big Data
Amasethi edatha amakhulu kakhulu adinga amathuluzi akhethekile ukuze agcine, ahlaziye, futhi akhiphe inani, avame ukusetshenziselwa ukuqeqesha amamodeli e-AI.
Isibonelo: Ukusebenzisa izigidi zokusebenzisana kwabasebenzisi ukuqeqesha izinjini zokuncoma zamapulatifomu e-e-commerce.
Imodeli Yebhokisi Elimnyama (Black Box Model)
Black Box Model
Uhlobo lwe-AI noma imodeli yokufunda komshini enengqondo yangaphakathi engahumushi kalula ngabantu, okwenza kube nzima ukuqonda ukuthi izinqumo zenziwa kanjani.
Isibonelo: Inethiwekhi ye-neural ejulile esetshenziselwa ukuvumela izikweletu kodwa engahlinzeki ngencazelo ecacile yokuthi kungani umfakisicelo oyedwa wamukelwe kanti omunye wenqatshwa.
Ukucubungula Kwengqondo (Cognitive Computing)
Cognitive Computing
Izinhlelo ze-AI eziklanyelwe ukulingisa izinqubo zokucabanga zomuntu, njengokucabanga nokufunda, zisebenzisa amasu afana ne-NLP nokuqashelwa kwamaphethini.
Isibonelo: Uhlelo lokucubungula kwengqondo olusiza ochwepheshe bezomthetho ukuhlaziya umthetho wecala nokubikezela imiphumela.
Umbono Wekhompyutha (Computer Vision)
Computer Vision
Insimu yobuhlakani bokwenziwa evumela amakhompyutha ukuba ahumushe futhi acubungule idatha ebonakalayo njengezithombe namavidiyo.
Isibonelo: Izinhlelo zokuqashelwa kobuso ezikhomba abantu kumavidiyo okuphepha zisebenzisa umbono wekhompyutha.
I-Corpus (Corpus)
Corpus
Iqoqo elikhulu lemibhalo ebhaliwe noma ekhulunywayo esetshenziselwa ukuqeqesha amamodeli olimi.
Isibonelo: Isethi yedatha ye-Common Crawl iyi-corpus yewebhu yomphakathi esetshenziselwa ukuqeqesha amamodeli olimi amakhulu njenge-GPT.
Ukuhamba Kwedatha (Data Drift)
Data Drift
Isenzakalo lapho idatha yokufaka ishintsha ngokuhamba kwesikhathi, okwenza ukusebenza kwemodeli kwehle.
Isibonelo: Imodeli yokulungisa ebikezelayo yemishini yezimboni iba ngaphansi kokunembile njengoba kungeniswa ubuchwepheshe obusha bezinzwa.
Ukulebula Idatha (Data Labelling)
Data Labelling
Inqubo yokufaka amanothi edatha ngamathegi noma amalebula ukuze ilungele ukufunda okugadiwe.
Isibonelo: Ukulebula izinkulungwane zezithombe zesimila njengezingalimazi noma ezinomdlavuza ukuqeqesha imodeli yokuthola umdlavuza.
Ukumbiwa Kwedatha (Data Mining)
Data Mining
Inqubo yokuthola amaphethini anenjongo, ukuhlobana, kanye nokungajwayelekile kumasethi edatha amakhulu.
Isibonelo: Abathengisi abasebenzisa ukumbiwa kwedatha ukuthola ukuthi abantu abathenga amanabukeni bavame ukuthenga ubhiya.
Ukufunda Okujulile (Deep Learning)
Deep Learning
Insimu encane yokufunda komshini esebenzisa amanethiwekhi e-neural anamalayeri amaningi ukwenza amaphethini ayinkimbinkimbi edatha.
Isibonelo: Ukufunda okujulile kusetshenziswa kumamodeli olimi njenge-GPT-4 namanye amamodeli okukhiqiza izithombe njenge-Stable Diffusion.
Amamodeli Okusabalalisa (Diffusion Models)
Diffusion Models
Uhlobo lwamamodeli okukhiqiza afunda ukukhiqiza idatha ngokuguqula kancane kancane umsindo ongahleliwe ube imiphumela ehlelekile.
Isibonelo: I-Stable Diffusion idala izithombe ezingokoqobo kusuka kumibhalo esebenzisa amasu okusabalalisa.
Ukufaka (Embedding)
Embedding
Ukumelela okuyinombolo kwedatha, okuvame ukusetshenziselwa ukuthatha incazelo yesemantic yamagama, izithombe, noma imisho.
Isibonelo: Ku-NLP, igama elithi 'ibhange' lingaba nokufaka okufanayo 'nemali' kodwa okuhlukile 'ebhange lomfula' kuye ngomongo.
I-Epoch (Epoch)
Epoch
Ukuphindaphinda okuphelele kuyo yonke isethi yedatha yokuqeqesha ngesikhathi senqubo yokuqeqesha imodeli yokufunda komshini.
Isibonelo: Uma isethi yedatha inezibonelo eziyi-1,000 futhi imodeli ibona zonke kanye ngesikhathi sokuqeqesha, lokho kuyi-epoch eyodwa.
I-AI Yokuziphatha (Ethical AI)
Ethical AI
Ifilosofi yokuklama nokusebenzisa eqinisekisa ukuthi ubuchwepheshe be-AI busebenza ngokusobala, ngokulingana, futhi ngokuhambisana nezindinganiso zomphakathi.
Isibonelo: Ithuluzi lokuqasha le-AI elihlanganisa ukuhlolwa kokukhetha ukuvimbela ukubandlulula abantu abancane.
Uhlelo Lwezingcweti (Expert System)
Expert System
Uhlelo lwe-AI olulingisa amakhono okwenza izinqumo ochwepheshe bomuntu endaweni ethile esebenzisa imithetho nengqondo.
Isibonelo: Uhlelo lwezingcweti olusetshenziswa kwezolimo ukuncoma ukwelashwa kwezitshalo ngokusekelwe kudatha yenhlabathi nomlando wezinambuzane.
I-AI Echazekayo (XAI) (Explainable AI (XAI))
Explainable AI (XAI)
Izinhlelo ze-AI eziklanyelwe ukwenza izinqubo zazo zangaphakathi nezinqumo ziqondakale kubantu, zikhulisa ukwethenjwa nokuziphendulela.
Isibonelo: I-AI yokuxilonga yezokwelapha engahlinzeki nje kuphela ngesincomo kodwa futhi ichaza ukuthi yiziphi izimpawu eziholele kuleso siphetho.
Ukufunda Okuncane (Few-shot Learning)
Few-shot Learning
Indlela yokufunda komshini lapho imodeli iqeqeshwa noma ilungiswa kahle isebenzisa kuphela inani elincane lezibonelo ezilebuliwe.
Isibonelo: Ukwenza i-LLM ngokwezifiso ukubhala ama-imeyili asemthethweni ngemuva kokuyibonisa izibonelo eziyi-10 kuphela.
Ukulungisa Okuncane (Fine-tuning)
Fine-tuning
Inqubo yokuthatha imodeli esiqeqeshwe ngaphambili nokuyiqeqesha ngokuqhubekayo kusethi yedatha entsha, encane ukuze iyenze ikhetheke emsebenzini othile.
Isibonelo: Ukulungisa kahle i-LLM ejwayelekile njenge-GPT kumadokhumenti asemthethweni angaphakathi ukuze kwakhiwe umsizi wokubhala umthetho.
Imodeli Yesisekelo (Foundation Model)
Foundation Model
Imodeli enkulu eqeqeshwe kudatha ehlukahlukene futhi ebanzi engaguqulelwa emisebenzini eminingi elandelayo.
Isibonelo: I-GPT-4 ne-PaLM 2 amamodeli esisekelo akwazi ukufingqa, i-Q&A, ukuhumusha, nokuningi.
I-Fuzzy Logic (Fuzzy Logic)
Fuzzy Logic
Uhlobo lwengqondo olubhekana namanani alinganiselwe esikhundleni sengqondo eqinile ye-true/false (binary), ewusizo ekucabangeni ngaphansi kokungaqiniseki.
Isibonelo: Isetshenziswa ezinhlelweni zokulawula isimo sezulu ukulungisa izinga lokushisa ngokusekelwe kokufaka okungacacile njengokuthi 'kushisa kancane' noma 'kubanda kakhulu'.
Inethiwekhi Yokulwa Yokukhiqiza (GAN) (Generative Adversarial Network (GAN))
Generative Adversarial Network (GAN)
Ukwakhiwa kwemodeli yokukhiqiza lapho amanethiwekhi amabili — umkhiqizi nombandlululi — eqhudelana ukuze athuthukise ikhwalithi yokuphumayo.
Isibonelo: Ama-GAN asetshenziselwa ukudala amavidiyo e-deepfake noma ukukhiqiza izithombe zomkhiqizo ezingokoqobo kusuka kumidwebo.
I-AI Yokukhiqiza (Generative AI)
Generative AI
Isigaba sobuhlakani bokwenziwa esingakhiqiza okuqukethwe okusha — njengombhalo, izithombe, umculo, noma ividiyo — kusuka kudatha yokuqeqesha.
Isibonelo: I-ChatGPT ekhiqiza okuthunyelwe kubhulogi noma i-Midjourney edala ubuciko bedijithali kusuka kumibhalo.
I-Transformer Yokukhiqiza Efundiswe Ngaphambili (GPT) (Generative Pre-trained Transformer (GPT))
Generative Pre-trained Transformer (GPT)
Uhlobo lwamamodeli olimi amakhulu athuthukiswe yi-OpenAI esebenzisa ukwakhiwa kwe-transformer futhi aqeqeshwe ngaphambili ngamanani amakhulu edatha yombhalo ukuze enze imisebenzi ehlukahlukene yolimi.
Isibonelo: I-GPT-4 iyakwazi ukubhala izindatshana, ukuhumusha izilimi, nokufingqa amadokhumenti ngesikhuthazo esincane.
I-Genetic Algorithm (Genetic Algorithm)
Genetic Algorithm
Indlela yokulungisa ethonywe ukukhethwa kwemvelo lapho izixazululo zithuthuka ngokuhamba kwesikhathi ngokuguqulwa, ukuwela, nokukhethwa.
Isibonelo: Isetshenziselwa ukuklama ukwakhiwa kwenethiwekhi ye-neural esebenzayo ngokulingisa ukusinda kwabafaneleka kakhulu.
Ukukhohlisa (Hallucination)
Hallucination
Ukukhiqizwa kokuqukethwe okuzwakala kunengqondo kodwa okungelona iqiniso noma okungenangqondo ngemodeli ye-AI.
Isibonelo: Imodeli yolimi ekhiqiza isitatimende esingekho noma inikeza amaqiniso omlando angamanga.
I-Heuristic (Heuristic)
Heuristic
Indlela esebenzayo yokuxazulula izinkinga engakuqinisekisi isixazululo esiphelele kodwa eyanele ngezinjongo ezisheshayo.
Isibonelo: Ukusebenzisa umthetho wesithupha ukulinganisa isikhathi sokulethwa ohlelweni lwe-AI lokuhambisa izimpahla.
I-Hyperparameter (Hyperparameter)
Hyperparameter
Inani lokulungiselela elisethwe ngaphambi kokuqeqesha imodeli yokufunda komshini, njengezinga lokufunda noma inani lamalayeri.
Isibonelo: Ukulungisa usayizi we-batch kusuka ku-32 kuya ku-128 ukuze kuthuthukiswe isivinini sokuqeqesha nokusebenza kwemodeli.
Ukuphetha (Inference)
Inference
Inqubo yokusebenzisa imodeli yokufunda komshini eqeqeshwe ukuze yenze izibikezelo noma ikhiqize imiphumela kusuka kudatha entsha yokufaka.
Isibonelo: Ukusebenzisa imodeli ye-GPT elungiswe kahle ukubhala ama-imeyili eqembu lokusekela amakhasimende.
Ukuthola Inhloso (Intent Detection)
Intent Detection
Umsebenzi ekuqondeni ulimi lwemvelo lapho uhlelo luhlonza inhloso noma injongo yomsebenzisi emlayezweni.
Isibonelo: Ku-chatbot, ukuqashela 'Ngifuna ukubhuka indiza' njengenhloso yokubhuka uhambo.
I-Internet Yezinto (IoT) (Internet of Things (IoT))
Internet of Things (IoT)
Inethiwekhi yamadivayisi aphathekayo axhunyiwe afakwe izinzwa, isofthiwe, namanye ubuchwepheshe ukuqoqa nokushintshisana ngedatha.
Isibonelo: Ama-thermostat ahlakaniphile namafriji abika idatha yokusebenzisa futhi alungise izilungiselelo esebenzisa ukuhlaziya kwe-AI.
Ukuhumusha (Interpretability)
Interpretability
Izinga lapho umuntu angaqonda khona izindlela zangaphakathi zemodeli yokufunda komshini kanye nenqubo yayo yokwenza izinqumo.
Isibonelo: Isihlahla sokwenza izinqumo sihumushwa kalula kunenethiwekhi ye-neural ejulile ngoba izinqumo zayo zingalandeleka.
I-Jupyter Notebook (Jupyter Notebook)
Jupyter Notebook
Indawo yokusebenza yekhompyutha evulekile evumela abasebenzisi ukuba babhale ikhodi, babonise imiphumela, futhi babhale ukuhlaziya kusixhumanisi esisodwa.
Isibonelo: Ososayensi bedatha basebenzisa ama-Jupyter Notebooks ukwenza amamodeli wokufunda komshini futhi babelane ngemiphumela.
I-K-Nearest Neighbours (KNN) (K-Nearest Neighbours (KNN))
K-Nearest Neighbours (KNN)
I-algorithm yokufunda komshini elula, engahleliwe esetshenziselwa ukuhlukanisa nokubuyisela emuva. Yenza izinqumo ngokusekelwe ezibonelweni zokuqeqesha eziseduze endaweni yesici.
Isibonelo: Ukuze uhlukanise isithelo esisha njenge-apula noma ipheya, i-KNN ihlola ukuthi yiziphi izithelo ezilebuliwe eziseduze ngesimo nombala.
I-Knowledge Graph (Knowledge Graph)
Knowledge Graph
Ukwakhiwa kwedatha okusebenzisa ama-node nama-edge ukumelela nokugcina izincazelo ezixhunyiwe zezinto kanye nobudlelwane bazo.
Isibonelo: Iphaneli yolwazi ye-Google inikwe amandla yi-knowledge graph exhuma izinto ezifana nabantu, izindawo, nemicimbi.
Ukulungiswa Kwemodeli Yokufunda Ulimi (LLMO) (Language Learning Model Optimisation (LLMO))
Language Learning Model Optimisation (LLMO)
Amasu asetshenziselwa ukuthuthukisa ukusebenza, ukusebenza kahle, noma ukuguquguquka kwamamodeli olimi amakhulu emisebenzini ethile noma ezizindeni.
Isibonelo: Ukusebenzisa ukulinganisa nokulungisa imiyalelo ukuze kulungiswe i-LLM ukuze isetshenziswe ebhizinisini.
Imodeli Yolimi Olukhulu (LLM) (Large Language Model (LLM))
Large Language Model (LLM)
Uhlobo lwemodeli yokufunda ejulile eqeqeshwe ngamanani amakhulu edatha yombhalo ekwazi ukukhiqiza, ukuqonda, nokucabanga ngolimi lomuntu.
Isibonelo: I-ChatGPT ne-Claude ama-LLM aqeqeshwe ukusiza ekubhaleni, ekubhaleni ikhodi, nasekuphenduleni imibuzo.
Indawo Efihliwe (Latent Space)
Latent Space
Ukumelela okungabonakali okunamandla amakhulu lapho okufaka okufanayo kuhlelwa khona eduze, okusetshenziswa kumamodeli okukhiqiza nokufaka.
Isibonelo: Ekukhiqizeni izithombe, ukuguqula indawo efihliwe kungashintsha izici ezifana nokukhanya noma imizwa.
Izinga Lokufunda (Learning Rate)
Learning Rate
I-hyperparameter ebalulekile ekuqeqesheni elawula ukuthi izisindo zemodeli zilungiswa kangakanani ngokuphathelene ne-loss gradient.
Isibonelo: Izinga lokufunda eliphezulu lingaholela ekudluleni okuncane, kanti izinga eliphansi kakhulu linciphisa inqubekelaphambili yokuqeqesha.
Ukufunda Komshini (ML) (Machine Learning (ML))
Machine Learning (ML)
Igatsha le-AI elivumela izinhlelo ukuba zifunde kudatha futhi zithuthukise ukusebenza ngaphandle kokuhlelwa ngokusobala.
Isibonelo: Izihlungi ze-spam zisebenzisa ukufunda komshini ukuhlukanisa ama-imeyili njenge-spam noma cha ngokusekelwe ezibonelweni ezidlule.
Ukuhamba Kwemodeli (Model Drift)
Model Drift
Isenzakalo lapho ukunemba kwemodeli kwehle ngokuhamba kwesikhathi ngenxa yezinguquko kudatha noma endaweni.
Isibonelo: Imodeli yokuthola ukukhwabanisa iba ngaphansi kokunembile njengoba amasu okukhwabanisa ethuthuka.
Ukuqeqeshwa Kwemodeli (Model Training)
Model Training
Inqubo yokufaka idatha kumodeli yokufunda komshini nokulungisa amapharamitha ayo ukuze kuncishiswe iphutha.
Isibonelo: Ukuqeqesha injini yokuncoma emlandweni wokuthenga kwamakhasimende ukuze kunconywe imikhiqizo emisha.
I-AI Yezinhlobo Eziningi (Multimodal AI)
Multimodal AI
Izinhlelo ze-AI ezikwazi ukucubungula nokuhlanganisa izinhlobo eziningi zedatha njengombhalo, izithombe, umsindo, namavidiyo.
Isibonelo: Imodeli efana ne-GPT-4 Vision engakwazi ukufunda umbhalo futhi ihumushe izithombe ngesikhathi esifanayo.
Ukucubungula Ulimi Lwemvelo (NLP) (Natural Language Processing (NLP))
Natural Language Processing (NLP)
Insimu encane ye-AI egxile ekusebenzisaneni phakathi kwamakhompyutha nezilimi zabantu (zemvelo). Ivumela imishini ukuba ifunde, iqonde, futhi iphendule ngolimi lomuntu.
Isibonelo: I-NLP isetshenziswa kubasizi bezwi, izinhlelo zokuhumusha ulimi, nama-chatbot.
Inethiwekhi Ye-Neural (Neural Network)
Neural Network
Imodeli yokufunda komshini ethonywe ukwakhiwa kobuchopho bomuntu, eyakhiwe ngamalayeri ama-node axhunyiwe (ama-neuron).
Isibonelo: Amanethiwekhi e-neural angemuva kwamamodeli okufunda ajulile asetshenziswa ekuqashelweni kwezithombe nezinkulumo.
Umsindo (Noise)
Noise
Ulwazi olungahleliwe noma olungahlobene nedatha olungafihla amaphethini anenjongo futhi luthinte kabi ukusebenza kwemodeli.
Isibonelo: Amaphutha enzwa noma okufaka idatha okugcwele amaphutha kungabhekwa njengomsindo.
I-Ontology (Ontology)
Ontology
Uhlaka oluhlelekile oluhlukanisa futhi luchaze ubudlelwane phakathi kwemiqondo ngaphakathi kwesizinda, okuvame ukusetshenziswa ezinhlelweni ze-AI zesemantic.
Isibonelo: I-ontology kwezokunakekelwa kwempilo ingachaza ukuthi izimpawu zihlobana kanjani nezifo nokwelashwa.
Ukufaka Ngokweqile (Overfitting)
Overfitting
Iphutha lokwenza imodeli lapho imodeli yokufunda komshini ithatha umsindo kudatha yokuqeqesha futhi yenza kabi kudatha entsha.
Isibonelo: Imodeli ebamba ngekhanda izimpendulo zokuqeqesha kodwa engakwazi ukuphatha idatha yokuhlola engabonwa ifakwe ngokweqile.
Ukuhlaziya Okubikezelayo (Predictive Analytics)
Predictive Analytics
Ukusetshenziswa kwedatha, ama-algorithms, kanye ne-AI ukuhlonza amathuba emiphumela yesikhathi esizayo ngokusekelwe kudatha yomlando.
Isibonelo: Abathengisi basebenzisa ukuhlaziya okubikezelayo ukuze babikezele isidingo semikhiqizo ethile.
Ukuqeqeshwa Ngaphambili (Pre-training)
Pre-training
Inqubo yokuqeqesha imodeli ekuqaleni kusethi yedatha enkulu, ejwayelekile ngaphambi kokuyilungisa kahle emisebenzini ethile.
Isibonelo: Amamodeli e-GPT aqeqeshwe ngaphambili kuma-corpora amakhulu ngaphambi kokwenziwa ngokwezifiso kuma-chatbot okusekela amakhasimende.
Ubunjiniyela Bokukhuthaza (Prompt Engineering)
Prompt Engineering
Ubuciko nesayensi yokwakha izikhuthazo ezisebenzayo ukuze kuqondiswe ukuphumayo kwamamodeli olimi amakhulu.
Isibonelo: Ukwengeza imiyalelo yesistimu efana nokuthi 'Phendula njengomfundisi ohloniphekile' kuyisibonelo sobunjiniyela bokukhuthaza.
Ukulinganisa (Quantisation)
Quantisation
Indlela yokucindezela imodeli enciphisa inani lamabhithi asetshenziselwa ukumelela izisindo nokusebenza, kuthuthukisa ukusebenza kahle.
Isibonelo: Ukulinganisa imodeli kusuka ku-32-bit kuya ku-8-bit kuthuthukisa ukusebenza kumadivayisi eselula.
I-Quantum Computing (Quantum Computing)
Quantum Computing
Indlela entsha yokucubungula esekelwe kumakhenikha e-quantum, ephethe amandla okucubungula okukhulu.
Isibonelo: I-Quantum computing ingase ngelinye ilanga isheshise ukuqeqeshwa kwe-AI ngaphezu kwemingcele yakudala.
Injini Yokucabanga (Reasoning Engine)
Reasoning Engine
Uhlelo ku-AI oluthola iziphetho ezinengqondo kusuka eqoqweni lamaqiniso noma idatha esebenzisa imithetho noma ama-algorithms okuphetha.
Isibonelo: Ithuluzi lokuxilonga le-AI lisebenzisa injini yokucabanga ukuze lithole izimo zezokwelapha ezingenzeka ngokusekelwe ezimpawu.
Ukufunda Ngokuqinisa (RL) (Reinforcement Learning (RL))
Reinforcement Learning (RL)
Insimu yokufunda komshini lapho ama-agent efunda ngokusebenzisana nendawo yawo ukuze akhulise imivuzo eqoqiwe.
Isibonelo: Irobhothi elifunda ukuhamba ngokuzama nangamaphutha lisebenzisa amasu e-RL.
Ukufunda Ngokuqinisa Ngempendulo Yomuntu (RLHF) (Reinforcement Learning with Human Feedback (RLHF))
Reinforcement Learning with Human Feedback (RLHF)
Indlela yokufunda lapho izintandokazi zabantu ziqondisa isignali yomvuzo ye-AI, okuvame ukusetshenziswa ekulungiseni kahle amamodeli olimi.
Isibonelo: I-ChatGPT yaqeqeshwa nge-RLHF ukuze ikhiqize izimpendulo eziwusizo nezilula.
Ukukhiqiza Okukhulisiwe Ngokubuyisa (RAG) (Retrieval-Augmented Generation (RAG))
Retrieval-Augmented Generation (RAG)
Indlela ehlanganisa ukubuyiswa kolwazi nokukhiqiza, lapho i-LLM ithola amadokhumenti ahlobene ukuze ithuthukise impendulo yayo.
Isibonelo: Umsizi we-AI uthola futhi acaphune izincazelo zomkhiqizo ngenkathi ekhiqiza impendulo kumbuzo wobuchwepheshe.
Ukufunda Okuzimele (Self-Supervised Learning)
Self-Supervised Learning
Indlela yokuqeqesha lapho imodeli ifunda amaphethini ngokukhiqiza amalebula ayo kusuka kudatha eluhlaza, kunciphisa ukuthembela kudatha efakwe amanothi ngabantu.
Isibonelo: I-BERT iqeqeshwa ngokufunda okuzimele ngokubikezela amagama alahlekile embhalweni.
Ukusesha Okunengqondo (Semantic Search)
Semantic Search
Indlela yokusesha eqonda inhloso yomsebenzisi nencazelo yomongo, hhayi nje ukufanisa amagama angukhiye.
Isibonelo: Ukusesha 'indlela yokulungisa umpompi ovuzayo' kubuyisa imihlahlandlela ngisho noma igama elithi 'umpompi ovuzayo' lingekho kudokhumenti.
Ukuhlaziya Imizwa (Sentiment Analysis)
Sentiment Analysis
Inqubo yokuhlonza imizwa, imibono, noma izimo zengqondo embhalweni, okuvame ukuhlukaniswa njengokuhle, okubi, noma okungathathi hlangothi.
Isibonelo: Ukuhlaziya ama-tweet ukuze kulinganiswe ukusabela komphakathi kumkhiqizo omusha.
I-Stochastic (Stochastic)
Stochastic
Okuhlanganisa ukungahleliwe noma ukuziphatha okungenzeka, okuvame ukusetshenziswa ku-AI yokukhiqiza nama-algorithms okulungisa.
Isibonelo: Ukuphumayo kwe-GPT-4 kuyahlukahluka kokufaka okufanayo ngenxa yenqubo yayo yokuchaza engahleliwe.
I-Strong AI (Strong AI)
Strong AI
Futhi eyaziwa ngokuthi i-Artificial General Intelligence (AGI), ibhekisela emishinini enamakhono okucabanga asezingeni lomuntu kuzo zonke izizinda.
Isibonelo: I-AI yesikhathi esizayo engakwazi ukubhala amanoveli ngokuzimela, ihlele amadolobha, futhi ixazulule izinkinga zokuziphatha ngokulinganayo.
I-Super Artificial Intelligence (SAI) (Super Artificial Intelligence (SAI))
Super Artificial Intelligence (SAI)
I-AI eyimfundiso edlula kakhulu ubuhlakani bomuntu kuzo zonke izici—ukucabanga, ubuhlakani, ubuhlakani bemizwa, njll.
Isibonelo: I-SAI ingase ngelinye ilanga ithuthukise isayensi entsha namafilosofi ngokuzimela.
Ukufunda Okugadiwe (Supervised Learning)
Supervised Learning
Indlela yokufunda komshini lapho amamodeli eqeqeshwa kudatha elebuliwe ukuze afunde ukufaka-ukuphumayo.
Isibonelo: Ukufundisa imodeli ukuhlukanisa ama-imeyili njenge-spam noma cha isebenzisa izibonelo zomlando.
Idatha Yokwenziwa (Synthetic Data)
Synthetic Data
Idatha ekhiqizwe ngokwenziwa elingisa idatha yangempela, okuvame ukusetshenziselwa ukuqeqesha lapho idatha yangempela incane noma ibucayi.
Isibonelo: Ukudala izithombe zezokwelapha zokwenziwa ukuqeqesha amamodeli okuxilonga ngaphandle kokwephula ubumfihlo besiguli.
Ithokheni (Token)
Token
Iyunithi yombhalo ecubungulwa ama-LLM—ngokuvamile igama noma ingxenye yegama.
Isibonelo: Umusho 'Sawubona mhlaba!' uhlukaniswa ube amathokheni ama-3: 'Sawubona', 'mhlaba', kanye '!'.
Ukwenza Amathokheni (Tokenisation)
Tokenisation
Inqubo yokuhlukanisa umbhalo ube amathokheni ukuze ucubungulwe imodeli.
Isibonelo: Ku-NLP, 'I-ChatGPT inhle' iba ['Chat', 'G', 'PT', 'inhle'].
Ukufunda Ngokudlulisa (Transfer Learning)
Transfer Learning
Ukusetshenziswa kolwazi oluvela emsebenzini owodwa ukuze kuthuthukiswe ukufunda komunye umsebenzi ohlobene, kunciphisa isikhathi sokuqeqesha nezidingo zedatha.
Isibonelo: Ukulungisa kahle imodeli eqeqeshwe embhalweni wesiNgisi ukuze yenze ukuhlaziya imizwa kolunye ulimi.
I-Transformer (Transformer)
Transformer
Ukwakhiwa kwenethiwekhi ye-neural esebenzisa izindlela zokunaka ukuze kwenziwe idatha elandelanayo, esetshenziswa kakhulu kuma-LLM.
Isibonelo: I-BERT, i-GPT, ne-T5 zonke zingamamodeli asekelwe ku-transformer.
Ukufaka Ngaphansi (Underfitting)
Underfitting
Lapho imodeli ilula kakhulu ukuthatha amaphethini kudatha yokuqeqesha, okuholela ekusebenzeni okubi.
Isibonelo: Imodeli elula ezama ukubikezela ukuhlukaniswa kwezithombe eziyinkimbinkimbi ingase ifake ngaphansi.
Ukufunda Okungagadiwe (Unsupervised Learning)
Unsupervised Learning
Indlela yokufunda lapho amamodeli ehlonza amaphethini noma amaqoqo kudatha engalebuliwe.
Isibonelo: Ukuhlanganisa amakhasimende ngokusekelwe ekuziphatheni kokuthenga ngaphandle kwamalebula achazwe ngaphambili.
Inhloso Yomsebenzisi (User Intent)
User Intent
Inhloso noma injongo engemuva kombuzo noma ukusebenzisana komsebenzisi.
Isibonelo: Umsebenzisi obhala 'indlela yokubhaka ikhekhe' cishe uhlose ukuthola iresiphi.
Isethi Yokuqinisekisa (Validation Set)
Validation Set
Ingxenye yedatha esetshenziselwa ukuhlola ukusebenza kwemodeli ngesikhathi sokuqeqesha nokulungisa ama-hyperparameter.
Isibonelo: Isetshenziselwa ukuthola ukufaka ngokweqile ngaphambi kokuhlolwa kokugcina.
I-Vector Database (Vector Database)
Vector Database
I-database eklanyelwe ukugcina nokusesha ukufaka kwe-vector okusetshenziswa emisebenzini ye-AI efana nokusesha okufanayo ne-RAG.
Isibonelo: I-Pinecone ne-Weaviate ama-vector database okugcina ukufaka umbhalo noma izithombe.
I-Vector Embedding (Vector Embedding)
Vector Embedding
Ukumelela okuyinombolo kwedatha okugcina incazelo yesemantic nobudlelwane endaweni ye-vector.
Isibonelo: Amagama 'inkosi' kanye 'nendlovukazi' anokufaka okufanayo nokuhlukana okuncane kobulili.
Umsizi Ongokoqobo (Virtual Assistant)
Virtual Assistant
Umenzeli wesofthiwe onikwe amandla yi-AI osiza abasebenzisi ukuqedela imisebenzi ngokuxoxa noma imiyalo yezwi.
Isibonelo: USiri, u-Alexa, noGoogle Assistant bangabasizi abathandwayo.
Ukuqashelwa Kwezwi (Voice Recognition)
Voice Recognition
Ubuchwepheshe obuhumusha futhi buguqule ulimi olukhulunywayo lube umbhalo noma isenzo.
Isibonelo: Ukubhala ngezwi nemiyalo yezwi kuncike ezinhlelweni zokuqashelwa kwezwi.
I-Weak AI (Weak AI)
Weak AI
Izinhlelo ze-AI eziklanyelwe ukwenza umsebenzi omncane, othile ngaphandle kobuhlakani obujwayelekile.
Isibonelo: I-AI edlala i-chess engakwazi ukuqonda ulimi noma ukushayela imoto iyisibonelo se-weak AI.
Ukukhipha Kuwebhu (Web Scraping)
Web Scraping
Ukukhishwa okuzenzakalelayo kolwazi kumawebhusayithi, okuvame ukusetshenziselwa ukuqoqa idatha yokuqeqesha noma ukuqapha okuqukethwe.
Isibonelo: Ukukhipha izikhangiso zezindlu ukuze kuqeqeshwe imodeli yokulinganisa impahla.
Isisindo (Weight)
Weight
Ipharamitha kumanethiwekhi e-neural elinquma amandla ethonya i-node eyodwa enalo kwenye.
Isibonelo: Izisindo zilungelelaniswa ngesikhathi sokuqeqesha ukuze kuncishiswe iphutha lemodeli.
I-Whisper (Whisper)
Whisper
Imodeli yokuguqula inkulumo ibe umbhalo eyathuthukiswa yi-OpenAI ekwazi ukuguqula umsindo ube umbhalo ngezilimi eziningi.
Isibonelo: I-Whisper ingaguqula izinkulumo nama-podcast ngokuqondile.
I-YAML (YAML)
YAML
Ifomethi efundeka kalula yokwenza idatha ibe uchungechunge, okuvame ukusetshenziselwa amafayela okulungiselela ezinhlelweni zokufunda komshini.
Isibonelo: Ukuchaza amapharamitha emodeli kufayela le-YAML ukuze kuqeqeshwe ku-PyTorch.
Ukufunda Okungenamuntu (Zero-shot Learning)
Zero-shot Learning
Ikhono lemodeli lokwenza imisebenzi engakaze iqeqeshwe ngokusobala ngokusebenzisa ulwazi olujwayelekile.
Isibonelo: Imodeli ephendula imibuzo yezomthetho naphezu kokungaqeqeshwa ngokuqondile kudatha yezomthetho.
I-Zettabyte (Zettabyte)
Zettabyte
Iyunithi yedatha yedijithali elingana ne-sextillion eyodwa (10^21) amabhayithi, okuvame ukusetshenziselwa ukuchaza ubukhulu bedatha ye-inthanethi.
Isibonelo: Ukuhamba kwe-inthanethi emhlabeni wonke kwadlula i-zettabyte eyodwa ngonyaka ngo-2016.
