Feere (Alignment)
Alignment
Kuugal ngam tabbintinde wonde faandaare, njeññi, e jikkuuji njuɓɓudi AI ina kawri e faandaare e nafoore aadee. Ɗum ina teeŋtini e njuɓɓudiiji ɓurɗi toowde ɗi mbaawi ƴellitde jikkuuji ɗi mbaɗaaka e anniya.
Misal: Tabbintinde wonde chatbot ngam wallitde cellal hakkille alaa ko waɗata ko bonnata hay so tawii ko e ɗaɓɓaande.
Kuutorgal Kuutorgal Kuutorgal (API) (Application Programming Interface (API))
Application Programming Interface (API)
Fedde sariyaaji e protokoluuji laaɓtuɗi ɗi mbaɗata njuɓɓudiiji software ceertuɗi mbaawa jokkondirde e waylude data.
Misal: Huutoraade OpenAI API ngam nelde ɗaɓɓaande e heɓde jaabawol ngol model ɗemngal waɗi e app web maa.
Hakkilantaagal Kuutorgal Keso (AGI) (Artificial General Intelligence (AGI))
Artificial General Intelligence (AGI)
No AI waÉ—irtee e miijooji É—i mbaawi waÉ—de golle hakkille kala É—e neÉ—É—o waawi waÉ—de. Ina janngina janngugol e nder fannuuji ceertuÉ—i.
Misal: Njuɓɓudi AGI ina waawi janngude waɗde jimɗi, waɗde opereeji, e waɗde jarribo filosofi tawa alaa ko waɗi e golle keertiiɗe.
Hakkilantaagal Kuutorgal (AI) (Artificial Intelligence (AI))
Artificial Intelligence (AI)
Simulaasiyoŋ hakkilantaagal aadee e nder masiŋaaji ɗi njuɓɓinaa ngam miijoo, miijoo, e golloo e hoore mum en.
Misal: AI ina wallita ballooɓe hoore mum en hono Siri e njuɓɓudiiji otooji hoore mum en hono Tesla Autopilot.
Etikaaji AI (AI Ethics)
AI Ethics
Fannu jokkondirÉ—o e batte etikaaji Æ´ellitaare e huutoraade AI, hawri e nuunÉ—al, sirlu, jaabawol, e luural.
Misal: WaÉ—de sariyaaji ngam haÉ—de algoritmuuji golle waÉ—de luural e dow jinnaaÉ—e walla leÆ´Æ´i.
Hakkilantaagal É“eydaangal (Augmented Intelligence)
Augmented Intelligence
Model jokkondiral É—o AI wallita e É“eyda hakkilantaagal aadee, tawa wonaa lomtaade É—um.
Misal: Kuutorgal radiyoloji ngal AI wallita ngal hollita ko boni e doktoor, mo waÉ—ata jaabawol sakkitiingol.
Kuutorgal Hoore mum (Autonomous Agent)
Autonomous Agent
Njuɓɓudi AI baawndi waɗde kuuje mum e waɗde golle ngam heɓde faandaare mum tawa alaa ko neɗɗo waɗi.
Misal: Robot neldugol hoore mum ina yahra e laabi wuro e haÉ—de ko boni e hoore mum.
Backpropagation (Backpropagation)
Backpropagation
Teknik ngam jannginde laylayɗe ƴiiƴam e waylude teddeendi gila e njeññi haa e laylayɗe naatgol, ngam ustude luural e anniyaaji.
Misal: Huutoraa e jannginde classifiers natal ngam ustude luural e anndude limÉ—e binndaaÉ—e e junngo.
Luural (Luural Algorithmic) (Bias (Algorithmic Bias))
Bias (Algorithmic Bias)
Yiɗde tawa alaa ko anniyaa e njeññi AI sabu data janngugol ngol alaa ko haani walla ngol alaa ko hollita.
Misal: Njuɓɓudi anndugol yeeso nde anndataa yimɓe ɓaleeɓe ko ɓuri heewde sabu ɓe ngalaa ko ɓuri heewde e data janngugol.
Data mawÉ—o (Big Data)
Big Data
Dataaji mawɗi no feewi ɗi njiɗi kuutorgal keertiiɗi ngam reende, ƴeewde, e heɓde nafoore, ko ɓuri heewde huutoraa ko ngam jannginde modeluuji AI.
Misal: Huutoraade miliyoŋaaji jokkondiral kuutorgal ngam jannginde engines recommendation ngam platforms e-commerce.
Model Boowal Baleewal (Black Box Model)
Black Box Model
No AI walla model janngugol masiŋ waɗirtee ɗo logic mum nder mum alaa ko neɗɗo waawi faamde, ko ɗum waɗi ina saɗi faamde no kuuje mbaɗirtee.
Misal: Laylayre ƴiiƴam toownde huutoraa ko ngam jaɓɓaade ñamaande kono alaa ko firti ko waɗi gooto jaɓɓaama goɗɗo jaɓɓaaka.
Hiisugol Hakkilantaagal (Cognitive Computing)
Cognitive Computing
Njuɓɓudiiji AI ɗi njuɓɓinaa ngam simulaade miijooji aadee, hono miijo e janngugol, huutoraade teknikuji hono NLP e anndugol mbaadi.
Misal: Njuɓɓudi hiisugol hakkilantaagal nde wallita karallaagal sariya ngam ƴeewde sariyaaji ñaawoore e anniyaade njeññi.
Yiyugol Kombiyuutaa (Computer Vision)
Computer Vision
Fannu hakkilantaagal kuutorgal ngal wallita kombiyuutaa ngam firtude e waÉ—de data yiyugol hono natal e wideyo.
Misal: Njuɓɓudiiji anndugol yeeso ɗi anndita yimɓe e nder wideyooji kisal huutoraade yiyugol kombiyuutaa.
Corpus (Corpus)
Corpus
Fedde mawnde binndi walla haalaaji huutoraaÉ—i ngam jannginde modeluuji É—emngal.
Misal: Dataset Common Crawl ko corpus web jamaa huutoraaÉ—o ngam jannginde modeluuji É—emngal mawÉ—i hono GPT.
Data Drift (Data Drift)
Data Drift
Fenomena É—o data naatgol waylata e nder sahaa, ko É—um waÉ—i model ina bonna.
Misal: Model maintenance anniyaade ngam kuutorgal industriyel ina ustoo no karallaagal sensor keso naatiraa.
Data Labelling (Data Labelling)
Data Labelling
Kuugal ngam waÉ—de data e tags walla labels ngam waÉ—de É—um haannde ngam janngugol supervised.
Misal: Waɗde labels e miliyoŋaaji natal tumor hono benign walla malignant ngam jannginde model anndugol kanseer.
Data Mining (Data Mining)
Data Mining
Kuugal ngam anndude mbaadiiji, jokkondiral, e ko boni e nder dataaji mawÉ—i.
Misal: Retailers huutoraade data mining ngam anndude wonde yimɓe ɓe coodata nappies ko ɓuri heewde coodata biya.
Janngugol Toowngol (Deep Learning)
Deep Learning
Fannu janngugol masiŋ ngal huutoraa ko laylayɗe ƴiiƴam keewɗe ngam waɗde mbaadiiji caɗtuɗi e nder data.
Misal: Janngugol toowngol ina huutoraa e modeluuji É—emngal hono GPT-4 e modeluuji waÉ—de natal hono Stable Diffusion.
Modeluuji Diffusion (Diffusion Models)
Diffusion Models
Fedde modeluuji generative ɗi janngata waɗde data e waylude sawta random e njeññi structured.
Misal: Stable Diffusion ina waÉ—a natal photorealistic gila e prompts binndaaÉ—i huutoraade teknikuji diffusion.
Embedding (Embedding)
Embedding
Firo vector limngal data, ko ɓuri heewde huutoraa ko ngam heɓde maana semantikaaji konnguɗi, natal, walla sentensi.
Misal: E nder NLP, konngol 'bank' ina waawi wonde e embeddings nannduÉ—i e 'money' kono ceertuÉ—i e 'riverbank' e dow kontekst.
Epoch (Epoch)
Epoch
Iteration timmunde e dow data janngugol timmungol e nder kuugal janngugol model janngugol masiŋ.
Misal: So dataset ina jogii 1,000 misal e model ina yiya É—i kala gooto e nder janngugol, É—um ko epoch gooto.
AI Etikaaji (Ethical AI)
Ethical AI
Filosofi design e deployment mo tabbintinta wonde karallaagal AI ina golloo e laaɓal, nuunɗal, e e dow nafoore renndo.
Misal: Kuutorgal golle AI ngal hawri e checks bias ngam haÉ—de luural e dow candidates minority.
Kuutorgal Kakkilantaagal (Expert System)
Expert System
Njuɓɓudi AI nde nanndiraa e baawɗe waɗde kuuje kakkilantaagal aadee e nder fannu keertiiɗo huutoraade sariyaaji e logic.
Misal: Kuutorgal kakkilantaagal huutoraa ko e ndema ngam jaɓɓaade safaaraaji ndema e dow data leydi e daartol bonannde.
AI Firoore (XAI) (Explainable AI (XAI))
Explainable AI (XAI)
Njuɓɓudiiji AI ɗi njuɓɓinaa ngam waɗde kuuje mum en nder e kuuje ɗe neɗɗo waawi faamde, ngam ɓeyda hoolaare e jaabawol.
Misal: AI diagnostic medikal ngal wonaa tan jaɓɓaade kono kadi firti ko ɗaɓɓaande nde waɗi ɗum.
Janngugol Few-shot (Few-shot Learning)
Few-shot Learning
No janngugol masiŋ waɗirtee ɗo model jannginaa walla fine-tuned huutoraade tan limɗe misaluuji labelled seeɗa.
Misal: WaÉ—de LLM ngam winndude emails sariya caggal nde hollitaa tan 10 misal.
Fine-tuning (Fine-tuning)
Fine-tuning
Kuugal ngam Æ´ettude model pre-trained e jannginde É—um kadi e dow dataset keso, tokooso ngam waÉ—de É—um keertiiÉ—o ngam golle keertiiÉ—e.
Misal: Fine-tuning LLM general hono GPT e dow documents sariya nder ngam waÉ—de balloowo drafting sariya.
Model Fondasiyoŋ (Foundation Model)
Foundation Model
Model mawÉ—o jannginaaÉ—o e dataaji ceertuÉ—i e yaajÉ—i É—i mbaawi wayleede ngam golle keewÉ—e.
Misal: GPT-4 e PaLM 2 ko modeluuji fondasiyoŋ baawɗi summarisation, Q&A, translation, e ko nanndi.
Fuzzy Logic (Fuzzy Logic)
Fuzzy Logic
No logic waɗirtee nde jokkondirta e limɗe approximate tawa wonaa logic true/false (binary) laaɓtuɗo, nafoore ngam miijo e nder luural.
Misal: Huutoraa ko e njuɓɓudiiji climate control ngam waylude temperatuure e dow inputs fuzzy hono 'a bit hot' walla 'very cold'.
Generative Adversarial Network (GAN) (Generative Adversarial Network (GAN))
Generative Adversarial Network (GAN)
Architecture model generative ɗo networks ɗiɗi — generator e discriminator — ina haɓa ngam ɓeyda quality njeññi.
Misal: GANs ina huutoraa ko ngam waÉ—de wideyooji deepfake walla waÉ—de natal product goonga gila e sketches.
Generative AI (Generative AI)
Generative AI
Fedde hakkilantaagal kuutorgal ngal waawi waɗde ko keso — hono binndol, natal, jimɗi, walla wideyo — gila e data janngugol.
Misal: ChatGPT ina waÉ—a blog posts walla Midjourney ina waÉ—a artwork digital gila e prompts binndaaÉ—i.
Generative Pre-trained Transformer (GPT) (Generative Pre-trained Transformer (GPT))
Generative Pre-trained Transformer (GPT)
Fedde modeluuji É—emngal mawÉ—i É—i OpenAI Æ´elliti É—i huutoraa ko architecture transformer e pre-trained e dow data binndol mawngol ngam waÉ—de golle É—emngal ceertuÉ—e.
Misal: GPT-4 ina waawi winndude essays, firtude É—emÉ—e, e summarisaade documents e prompts seeÉ—a.
Algorithme Genetic (Genetic Algorithm)
Genetic Algorithm
Teknik optimisation mo nanndiraa e selection natural É—o solutions Æ´ellitata e nder sahaa e mutation, crossover, e selection.
Misal: Huutoraa ko ngam design architecture neural network efficient e simulaade survival of the fittest.
Hallucination (Hallucination)
Hallucination
WaÉ—de ko nanndi e goonga kono ko alaa ko firti walla ko alaa ko firti e model AI.
Misal: Model É—emngal ina waÉ—a citation mo alaa walla ina waÉ—a facts daartol goonga.
Heuristic (Heuristic)
Heuristic
No problem-solving waÉ—irtee nde alaa ko tabbintinta solution timmunde kono ko haani ngam faandaare jooni.
Misal: Huutoraade rule of thumb ngam anniyaade sahaa neldugol e nder njuɓɓudi logistics AI.
Hyperparameter (Hyperparameter)
Hyperparameter
Limngal configuration set ko adii janngugol model janngugol masiŋ, hono learning rate walla limɗe layers.
Misal: Waylude batch size gila 32 haa 128 ngam É“eyda speed janngugol e performance model.
Inference (Inference)
Inference
Kuugal ngam huutoraade model jannginaaɗo ngam waɗde anniyaaji walla waɗde njeññi gila e data naatgol keso.
Misal: Huutoraade model GPT fine-tuned ngam waÉ—de emails ngam team support customer.
Faamugol Faandaare (Intent Detection)
Intent Detection
Golle e nder faamugol ɗemngal natural ɗo njuɓɓudi anndita faandaare walla anniya kuutorgal e nder mesaas.
Misal: E nder chatbot, anndude 'Mi yiÉ—i book flight' hono intent booking travel.
Internet of Things (IoT) (Internet of Things (IoT))
Internet of Things (IoT)
Network kuutorgal fisik jokkondirɗo e sensors, software, e karallaagal goɗɗo ngam heɓde e waylude data.
Misal: Smart thermostats e fridges ɗi njaɓɓata data huutoraade e waylude settings huutoraade AI analytics.
Firoore (Interpretability)
Interpretability
No neɗɗo waawi faamde mechanics nder model janngugol masiŋ e kuugal waɗde kuuje.
Misal: Decision tree ina É“uri faamde ko É“uri laylayre Æ´iiÆ´am toownde sabu kuuje mum ina mbaawi reede.
Jupyter Notebook (Jupyter Notebook)
Jupyter Notebook
Kuutorgal hiisugol interactive open-source ngal wallita kuutorgal ngam winndude code, yiyde njeññi, e document analysis e nder interface gooto.
Misal: Data scientists ina huutora Jupyter Notebooks ngam prototype modeluuji janngugol masiŋ e share njeññi.
K-Nearest Neighbours (KNN) (K-Nearest Neighbours (KNN))
K-Nearest Neighbours (KNN)
Algorithme janngugol masiŋ simple, non-parametric huutoraaɗo ngam classification e regression. Ina waɗa kuuje e dow misaluuji janngugol ɓurɗi ɓadaade e nder feature space.
Misal: Ngam classify fruit keso hono apple walla pear, KNN ina check fruits labelled É—i É“uri É“adaade e mbaadi e colour.
Knowledge Graph (Knowledge Graph)
Knowledge Graph
Njuɓɓudi data nde huutoraa ko nodes e edges ngam hollude e reende firooji jokkondirɗi e entities e jokkondiral mum en.
Misal: Panel knowledge Google ina wallita ko knowledge graph mo jokkondirta entities hono yimɓe, nokkuuji, e events.
Language Learning Model Optimisation (LLMO) (Language Learning Model Optimisation (LLMO))
Language Learning Model Optimisation (LLMO)
Teknikuji huutoraaÉ—i ngam É“eyda performance, efficiency, walla adaptability modeluuji É—emngal mawÉ—i ngam golle walla fannuuji keertiiÉ—i.
Misal: Huutoraade quantisation e instruction tuning ngam optimise LLM ngam huutoraade e nder enterprise.
Large Language Model (LLM) (Large Language Model (LLM))
Large Language Model (LLM)
No model janngugol toowngol waÉ—irtee jannginaaÉ—o e data binndol mawngol baawÉ—o waÉ—de, faamde, e miijoo e É—emngal aadee.
Misal: ChatGPT e Claude ko LLMs jannginaaÉ—i ngam wallitde e winndude, coding, e jaabawol questions.
Latent Space (Latent Space)
Latent Space
Firo abstract high-dimensional É—o inputs nannduÉ—i ina kawra É“adaade, huutoraa ko e modeluuji generative e embeddings.
Misal: E nder image generation, waylude latent space ina waawi waylude features hono brightness walla emotion.
Learning Rate (Learning Rate)
Learning Rate
Hyperparameter teeŋtuɗo e nder janngugol mo controlta no teddeendi model waylata e dow loss gradient.
Misal: Learning rate toownde ina waawi waÉ—de overshooting minima, tawa learning rate toownde ina ustoo progress janngugol.
Janngugol Masiŋ (ML) (Machine Learning (ML))
Machine Learning (ML)
Fannu AI ngal wallita njuɓɓudiiji ngam janngude gila e data e ɓeyda performance tawa alaa ko njuɓɓinaa e laaɓal.
Misal: Spam filters ina huutora janngugol masiŋ ngam classify emails hono spam walla not e dow misaluuji ɓennuɗi.
Model Drift (Model Drift)
Model Drift
Fenomena É—o accuracy model ustoo e nder sahaa sabu waylude data walla environment.
Misal: Model fraud detection ina ustoo no tactics fraud Æ´ellitata.
Janngugol Model (Model Training)
Model Training
Kuugal ngam waɗde data e model janngugol masiŋ e waylude parameters mum ngam ustude luural.
Misal: Jannginde recommendation engine e dow daartol coodugol customer ngam jaɓɓaade products keso.
Multimodal AI (Multimodal AI)
Multimodal AI
Njuɓɓudiiji AI baawɗi waɗde e jokkondirde types data keewɗi hono binndol, natal, sawta, e wideyo.
Misal: Model hono GPT-4 Vision ngal waawi janngude binndol e firtude natal e sahaa gooto.
Natural Language Processing (NLP) (Natural Language Processing (NLP))
Natural Language Processing (NLP)
Fannu AI ngal teeŋtini e jokkondiral hakkunde kombiyuutaa e ɗemɗe aadee (natural). Ina wallita masiŋaaji ngam janngude, faamde, e jaabawol e ɗemngal aadee.
Misal: NLP ina huutoraa ko e voice assistants, apps translation É—emngal, e chatbots.
Laylayre Æ´iiÆ´am (Neural Network)
Neural Network
Model janngugol masiŋ mo nanndiraa e njuɓɓudi hakkille aadee, mo hawri e layers nodes jokkondirɗi (neurons).
Misal: LaylayÉ—e Æ´iiÆ´am ina ngondi e modeluuji janngugol toowÉ—i huutoraaÉ—i e anndugol natal e haala.
Sawta (Noise)
Noise
Kabaaru random walla ko alaa ko firti e nder data mo waawi suuɗde mbaadiiji teeŋtuɗi e bonnude performance model.
Misal: Errors sensor walla data entries typo-filled ina mbaawi wonde sawta.
Ontology (Ontology)
Ontology
Njuɓɓudi structured nde categorisa e firti jokkondiral hakkunde miijooji e nder fannu, ko ɓuri heewde huutoraa ko e njuɓɓudiiji AI semantikaaji.
Misal: Ontology e nder cellal ina waawi firtude no symptoms jokkondirta e diseases e treatments.
Overfitting (Overfitting)
Overfitting
Luural modelling ɗo model janngugol masiŋ ina heɓa sawta e nder data janngugol e waɗa ko boni e data keso.
Misal: Model mo memorisa answers janngugol kono mo waawaa waÉ—de data test mo yiyaaka ko overfitted.
Predictive Analytics (Predictive Analytics)
Predictive Analytics
Huutoraade data, algoritmuuji, e AI ngam anndude ko waawi wonde e njeññi garooji e dow data daartol.
Misal: Retailers ina huutora predictive analytics ngam anniyaade ɗaɓɓaande ngam products keertiiɗi.
Pre-training (Pre-training)
Pre-training
Kuugal ngam jannginde model e dow data mawngol, general ko adii fine-tuning É—um ngam golle keertiiÉ—e.
Misal: Modeluuji GPT ko pre-trained e dow corpora mawÉ—i ko adii nde customised ngam chatbots customer service.
Prompt Engineering (Prompt Engineering)
Prompt Engineering
Art e science waɗde prompts efficient ngam steer njeññi modeluuji ɗemngal mawɗi.
Misal: WaÉ—de instructions system hono 'Jaabawol hono tutor polite' ko misal prompt engineering.
Quantisation (Quantisation)
Quantisation
Teknik compression model mo ustoo limÉ—e bits huutoraaÉ—i ngam hollude teddeendi e activations, ngam É“eyda efficiency.
Misal: Quantising model gila 32-bit haa 8-bit ina É“eyda performance e dow mobile devices.
Quantum Computing (Quantum Computing)
Quantum Computing
Paradigm hiisugol keso e dow mechanics quantum, mo jogii potential ngam baawÉ—e processing exponential.
Misal: Quantum computing ina waawi É“eyda janngugol AI ko É“uri limits classical.
Reasoning Engine (Reasoning Engine)
Reasoning Engine
Njuɓɓudi e nder AI nde heɓata njeññi logic gila e fedde facts walla data huutoraade sariyaaji walla algoritmuuji inference.
Misal: Kuutorgal diagnostic AI ina huutora reasoning engine ngam deduce conditions medical possible e dow symptoms.
Reinforcement Learning (RL) (Reinforcement Learning (RL))
Reinforcement Learning (RL)
Fannu janngugol masiŋ ɗo agents janngata e jokkondirde e environment mum en ngam ɓeyda rewards cumulative.
Misal: Robot ina janngina yahde e trial and error huutoraade teknikuji RL.
Reinforcement Learning with Human Feedback (RLHF) (Reinforcement Learning with Human Feedback (RLHF))
Reinforcement Learning with Human Feedback (RLHF)
No janngugol waÉ—irtee É—o preferences aadee ina ardina signal reward AI, ko É“uri heewde huutoraa ko e fine-tuning modeluuji É—emngal.
Misal: ChatGPT jannginaama e RLHF ngam waÉ—de jaabawol É“urÉ—o wallitde e kisal.
Retrieval-Augmented Generation (RAG) (Retrieval-Augmented Generation (RAG))
Retrieval-Augmented Generation (RAG)
No jokkondiral information retrieval e generation waɗirtee, ɗo LLM ina heɓa documents haanɗi ngam ɓeyda jaabawol mum.
Misal: Balloowo AI ina heɓa e cite specs product tawa ina waɗa jaabawol ngam question karallaagal.
Janngugol Hoore mum (Self-Supervised Learning)
Self-Supervised Learning
No janngugol waÉ—irtee É—o model janngata mbaadiiji e waÉ—de labels mum gila e data raw, ngam ustude reliance e data human-annotated.
Misal: BERT jannginaama e self-supervised learning e anniyaade konnguÉ—i missing e nder binndol.
Semantic Search (Semantic Search)
Semantic Search
Teknik yiilugol mo faamta faandaare kuutorgal e maana kontekstual, wonaa tan keyword matching.
Misal: Yiilugol 'no feere tap leaking' ina jaabawol guides hay so tawii konngol 'tap leaking' alaa e nder document.
Sentiment Analysis (Sentiment Analysis)
Sentiment Analysis
Kuugal ngam anndude emotions, opinions, walla attitudes e nder binndol, ko É“uri heewde classify hono positive, negative, walla neutral.
Misal: Ƴeewde tweets ngam heɓde reaction jamaa e product keso.
Stochastic (Stochastic)
Stochastic
Hawri e randomness walla jikku probabilistic, ko É“uri heewde huutoraa ko e generative AI e algoritmuuji optimisation.
Misal: Njeññi GPT-4 ina ceerta ngam input gooto sabu kuugal decoding stochastic mum.
AI Semmbe (Strong AI)
Strong AI
Kadi anndiraa hono Hakkilantaagal Kuutorgal Keso (AGI), ina firti masiŋaaji ɗi jogii baawɗe hakkille aadee e nder fannuuji kala.
Misal: AI garoowo mo waawi winndude novels, plan cities, e solve dilemmas etikaaji e nanndiraa.
Super Artificial Intelligence (SAI) (Super Artificial Intelligence (SAI))
Super Artificial Intelligence (SAI)
AI miijoore nde ɓuri hakkilantaagal aadee e fannuuji kala — miijo, creativity, emotional intelligence, etc.
Misal: SAI ina waawi Æ´ellitde sciences e philosophies keso e hoore mum.
Janngugol Supervised (Supervised Learning)
Supervised Learning
Teknik janngugol masiŋ ɗo modeluuji jannginaa ko e data labelled ngam janngude mappings input-output.
Misal: Jannginde model ngam classify emails hono spam walla not huutoraade misaluuji daartol.
Synthetic Data (Synthetic Data)
Synthetic Data
Data waÉ—aaÉ—o e kuutorgal mo simula data goonga, ko É“uri heewde huutoraa ko ngam janngugol so data goonga alaa walla ko sensitive.
Misal: WaÉ—de natal medical synthetic ngam jannginde modeluuji diagnostic tawa bonnata privacy patient.
Token (Token)
Token
Unit binndol waɗaaɗo e LLMs — ko ɓuri heewde ko konngol walla feccere konngol.
Misal: Sentensi 'Hello world!' ina feccaa e 3 tokens: 'Hello', 'world', e '!'.
Tokenisation (Tokenisation)
Tokenisation
Kuugal ngam feccude binndol e tokens ngam waÉ—de É—um e model.
Misal: E nder NLP, 'ChatGPT is great' ina wonti ['Chat', 'G', 'PT', 'is', 'great'].
Transfer Learning (Transfer Learning)
Transfer Learning
Huutoraade anndugol gila e golle gooto ngam É“eyda janngugol e golle goÉ—É—o jokkondirÉ—o, ngam ustude sahaa janngugol e data needs.
Misal: Fine-tuning model jannginaaÉ—o e binndol Engele ngam waÉ—de sentiment analysis e É—emngal goÉ—É—o.
Transformer (Transformer)
Transformer
Architecture neural network nde huutoraa ko attention mechanisms ngam waÉ—de mbaadiiji data sequential, huutoraa ko e LLMs.
Misal: BERT, GPT, e T5 ko modeluuji transformer-based.
Underfitting (Underfitting)
Underfitting
So model ina simple no feewi ngam heɓde mbaadiiji e nder data janngugol, ko ɗum waɗi performance bonɗo.
Misal: Model linear mo anniyaade classifications natal caÉ—tuÉ—i ina waawi underfit.
Janngugol Unsupervised (Unsupervised Learning)
Unsupervised Learning
No janngugol waÉ—irtee É—o modeluuji anndita mbaadiiji walla clusters e nder data unlabelled.
Misal: Grouping customers e dow purchasing behaviour tawa alaa labels predefined.
Faandaare Kuutorgal (User Intent)
User Intent
Faandaare walla anniya caggal query walla jokkondiral kuutorgal.
Misal: Kuutorgal winndude 'no bake cake' ina waawi yiɗde heɓde recipe.
Validation Set (Validation Set)
Validation Set
Subset data huutoraaÉ—o ngam Æ´eewde performance model e nder janngugol e tune hyperparameters.
Misal: Huutoraa ko ngam anndude overfitting ko adii testing sakkitiingol.
Vector Database (Vector Database)
Vector Database
Database designaaÉ—o ngam reende e yiiloo vector embeddings huutoraaÉ—i e golle AI hono similarity search e RAG.
Misal: Pinecone e Weaviate ko vector databases ngam reende text walla image embeddings.
Vector Embedding (Vector Embedding)
Vector Embedding
Firo numeric data mo reeni maana semantikaaji e jokkondiral e nder vector space.
Misal: KonnguÉ—i 'king' e 'queen' ina jogii embeddings nannduÉ—i e differences gender seeÉ—a.
Balloowo Virtual (Virtual Assistant)
Virtual Assistant
Agent software mo AI wallita mo wallita kuutorgal ngam timminde golle e dow conversation walla voice commands.
Misal: Siri, Alexa, e Google Assistant ko virtual assistants anndaaÉ—i.
Anndugol Daande (Voice Recognition)
Voice Recognition
Karallaagal ngal firtata e waylata É—emngal haala e binndol walla golle.
Misal: Voice typing e voice commands ina rely e voice recognition systems.
AI Loownde (Weak AI)
Weak AI
Njuɓɓudiiji AI ɗi njuɓɓinaa ngam waɗde golle keertiiɗe, tawa alaa hakkilantaagal general.
Misal: AI chess-playing mo waawaa faamde É—emngal walla drive oto ko misal weak AI.
Web Scraping (Web Scraping)
Web Scraping
Extraction automated information gila e websites, ko ɓuri heewde huutoraa ko ngam heɓde data janngugol walla monitor content.
Misal: Scraping real estate listings ngam jannginde model valuation property.
Teddeendi (Weight)
Weight
Parameter e nder neural networks mo anndita semmbe influence node gooto e goÉ—É—o.
Misal: Teddeendi ina waylata e nder janngugol ngam ustude luural model.
Whisper (Whisper)
Whisper
Model speech-to-text Æ´ellitaaÉ—o e OpenAI baawÉ—o transcribde audio e É—emÉ—e keewÉ—e.
Misal: Whisper ina waawi transcribde lectures e podcasts e accuracy toownde.
YAML (YAML)
YAML
Format human-readable ngam data serialisation, ko É“uri heewde huutoraa ko ngam configuration files e nder machine learning workflows.
Misal: Defining model parameters e nder YAML file ngam janngugol e PyTorch.
Janngugol Zero-shot (Zero-shot Learning)
Zero-shot Learning
Baawɗe model ngam waɗde golle ɗe jannginaaka e laaɓal e huutoraade anndugol general.
Misal: Model ina jaabawol questions sariya hay so tawii jannginaaka e data sariya.
Zettabyte (Zettabyte)
Zettabyte
Unit data digital nannduÉ—o e sextillion gooto (10^21) bytes, ko É“uri heewde huutoraa ko ngam firtude scale data internet.
Misal: Traffic internet winndere ndee É“uri 1 zettabyte e hitaande 2016.
