Alignement (Alignment)
Alignment
Mokanda ya kosala ete mikano, biloko oyo ebimaka, mpe bizaleli ya système ya AI ekokana na mikano mpe motuya ya bato. Yango ezali na ntina mingi na ba systèmes ya likolo oyo ekoki kobimisa bizaleli oyo ezali te na mokano ya polele.
Ndakisa: Kokamwa ete chatbot mpo na lisalisi ya nzoto ya maladi ekoki te kopesa toli ya misala ya mabe ata soki ezali na ba prompts.
Application Programming Interface (API) (Application Programming Interface (API))
Application Programming Interface (API)
Ezaleli ya mibeko mpe ba protocoles oyo elimbolami oyo epesaka nzela na ba systèmes ya logiciel ndenge na ndenge kosolola mpe kobongola ba données.
Ndakisa: Kosalela OpenAI API mpo na kotinda prompt mpe kozwa eyano oyo ebimisami na modèle ya lokota na application na yo ya web.
Artificial General Intelligence (AGI) (Artificial General Intelligence (AGI))
Artificial General Intelligence (AGI)
Lolenge ya AI ya théorie oyo ekoki kosala mosala nyonso ya mayele oyo moto akoki kosala. Epanzi koyekola na ba domaines nyonso.
Ndakisa: Système ya AGI ekoki koyekola kosala miziki, kosala ba opérations, mpe koleka examen ya philosophie kozanga programmation ya sikisiki mpo na mosala.
Artificial Intelligence (AI) (Artificial Intelligence (AI))
Artificial Intelligence (AI)
Simulation ya mayele ya bomoto na ba masini oyo epesami programme mpo na kokanisa, kokanisa, mpe kosala na bango moko.
Ndakisa: AI epesaka nguya na ba assistants personnels lokola Siri mpe ba systèmes ya kolakisa motuka na bango moko lokola Tesla Autopilot.
AI Ethics (AI Ethics)
AI Ethics
Discipline oyo etali makambo ya etika ya bokeli mpe kosalela ya AI, ezali na bosembo, kobombama ya makambo ya moto, mokumba, mpe kozanga discrimination.
Ndakisa: Kokela mibeko mpo na kopekisa ba algorithmes ya kosala misala ya kobuka mibeko na kotalela lolenge ya mobali to ya mwasi to ekólo.
Augmented Intelligence (Augmented Intelligence)
Augmented Intelligence
Modèle ya kosala elongo epai AI ekokisaka mpe ebongisaka mayele ya bomoto na esika ya kozongisa yango.
Ndakisa: Ba outils ya radiologie oyo esalaka na AI oyo ezali kolakisa ba anomalies mpo na ba docteurs, oyo bazali kozwa diagnostic ya suka.
Autonomous Agent (Autonomous Agent)
Autonomous Agent
Système ya AI oyo ekoki kozwa bikateli na yango moko mpe kosala misala mpo na kokokisa mikano na yango kozanga ete moto asala.
Ndakisa: Robot ya kolakisa biloko na yango moko oyo ezali kotambola na babalabala ya engumba mpe kokima mikakatano na yango moko.
Backpropagation (Backpropagation)
Backpropagation
Tekiniki ya koyekola ba réseaux neuronaux na kobongola ba poids na sens inverse kobanda na ba couches ya sortie tii na ba couches ya entrée, kokitisa ba erreurs ya prédiction.
Ndakisa: Ekomami na koyekola ba classificateurs ya bililingi mpo na kokitisa taux ya erreur na koyeba ba chiffres oyo ekomami na loboko.
Bias (Algorithmic Bias) (Bias (Algorithmic Bias))
Bias (Algorithmic Bias)
Bopusi ya kozanga mokano mpe ya système na ba résultats ya AI mpo na ba données ya koyekola oyo ezali te na bosembo to oyo ezali te na bopesi.
Ndakisa: Système ya koyeba bilongi oyo ekoki te koyeba bato ya langi ya ndenge na ndenge mpo na kozanga bopesi na ba données ya koyekola.
Big Data (Big Data)
Big Data
Ba ensembles ya données ya monene mingi oyo esengaka ba outils ya sikisiki mpo na kobomba, kotala, mpe kozwa motuya, mbala mingi esalelamaka mpo na koyekola ba modèles ya AI.
Ndakisa: Kosalela ba millions ya ba interactions ya ba utilisateurs mpo na koyekola ba moteurs ya recommandation mpo na ba plateformes ya e-commerce.
Black Box Model (Black Box Model)
Black Box Model
Lolenge ya modèle ya AI to ya koyekola ya masini oyo logique na yango ya kati ezali te pɛtɛɛ mpo na bato koyeba, kosalaka ete ezala mpasi koyeba ndenge bikateli ezwamaka.
Ndakisa: Réseau neuronal ya mozindo oyo esalelamaka mpo na kondimisa ba prêts kasi epesaka te malimboli ya polele mpo na nini mokomi moko andimami mpe mosusu aboyami.
Cognitive Computing (Cognitive Computing)
Cognitive Computing
Ba systèmes ya AI oyo esalemi mpo na kosala simulation ya ba processus ya makanisi ya bomoto, lokola kokanisa mpe koyekola, kosalelaka ba tekiniki lokola NLP mpe koyeba ba modèles.
Ndakisa: Système ya cognitive computing oyo esalisaka ba professionnels ya mibeko kotala mibeko ya likambo mpe kokanisa ba résultats.
Computer Vision (Computer Vision)
Computer Vision
Domaine ya mayele ya kosala oyo epesaka ba ordinateurs nzela ya kotala mpe kosala ba données ya bililingi lokola bililingi mpe ba vidéos.
Ndakisa: Ba systèmes ya koyeba bilongi oyo eyebisaka bato na ba vidéos ya sécurité kosalelaka computer vision.
Corpus (Corpus)
Corpus
Lisolo ya minene ya ba textes ekomami to elobami oyo esalelamaka mpo na koyekola ba modèles ya lokota.
Ndakisa: Common Crawl dataset ezali corpus ya web ya bato nyonso oyo esalelamaka mpo na koyekola ba modèles ya lokota ya minene lokola GPT.
Data Drift (Data Drift)
Data Drift
Phénomène epai ba données ya entrée ebongwanaka na ntango, kosalaka ete performance ya modèle ekita.
Ndakisa: Modèle ya prédiction ya maintenance mpo na ba équipements ya industrie ekómaka moke ya précision lokola tekiniki ya ba capteurs ya sika ebimisami.
Data Labelling (Data Labelling)
Data Labelling
Mokanda ya kopesa ba données ba tags to ba labels mpo na kosala ete ebonga mpo na koyekola ya supervisé.
Ndakisa: Kopesa ba milliers ya bililingi ya tumeur ba labels lokola benign to malignant mpo na koyekola modèle ya koyeba cancer.
Data Mining (Data Mining)
Data Mining
Mokanda ya koyeba ba modèles ya motuya, ba corrélations, mpe ba anomalies na ba ensembles ya données ya minene.
Ndakisa: Ba détaillants bazali kosalela data mining mpo na koyeba ete bato oyo basombaka ba couches mbala mingi basombaka mpe masanga.
Deep Learning (Deep Learning)
Deep Learning
Sous-domaine ya koyekola ya masini oyo esalelaka ba réseaux neuronaux ya ba couches ebele mpo na kosala ba modèles ya ba modèles ya mikakatano na ba données.
Ndakisa: Deep learning esalelamaka na ba modèles ya lokota lokola GPT-4 mpe ba modèles ya génération ya bililingi lokola Stable Diffusion.
Diffusion Models (Diffusion Models)
Diffusion Models
Lolenge ya ba modèles génératifs oyo eyekolaka kobimisa ba données na kobongola malembe malembe bruit ya hasard na ba résultats structurés.
Ndakisa: Stable Diffusion ekelaka bililingi ya photorealistic kobanda na ba prompts ya texte kosalelaka ba tekiniki ya diffusion.
Embedding (Embedding)
Embedding
Représentation ya vecteur numérique ya ba données, mbala mingi esalelamaka mpo na kozwa motuya ya sémantique ya maloba, bililingi, to ba phrases.
Ndakisa: Na NLP, liloba 'bank' ekoki kozala na ba embeddings ya ndenge moko na 'money' kasi ya ndenge na ndenge na 'riverbank' na kotalela contexte.
Epoch (Epoch)
Epoch
Itération ya mobimba na ensemble ya données ya koyekola mobimba na ntango ya processus ya koyekola ya modèle ya koyekola ya masini.
Ndakisa: Soki ensemble ya données ezali na ba exemples 1 000 mpe modèle emoni yango nyonso mbala moko na ntango ya koyekola, yango ezali epoch moko.
Ethical AI (Ethical AI)
Ethical AI
Philosophie ya conception mpe ya kosalela oyo ekamwaka ete ba tekiniki ya AI esalaka na polele, na bosembo, mpe na boyokani na motuya ya bato.
Ndakisa: Outil ya kosala misala ya AI oyo ezali na ba contrôles ya bopusi mpo na kopekisa discrimination na ba candidats ya minorité.
Expert System (Expert System)
Expert System
Système ya AI oyo esalaka simulation ya mayele ya kozwa bikateli ya moto ya mayele na domaine ya sikisiki kosalelaka mibeko mpe logique.
Ndakisa: Système ya expert oyo esalelamaka na agriculture mpo na kopesa toli ya ba traitements ya bilanga na kotalela ba données ya mabele mpe histoire ya ba parasites.
Explainable AI (XAI) (Explainable AI (XAI))
Explainable AI (XAI)
Ba systèmes ya AI oyo esalemi mpo na kosala ete ba processus na yango ya kati mpe bikateli na yango ezala pɛtɛɛ mpo na bato koyeba, kokómisa confiance mpe mokumba.
Ndakisa: AI ya diagnostic médical oyo epesaka te kaka recommandation kasi mpe elimbolaka ba symptômes nini eleki na conclusion yango.
Few-shot Learning (Few-shot Learning)
Few-shot Learning
Méthode ya koyekola ya masini epai modèle eyekolami to ebongisami na kosalelaka kaka motango moke ya ba exemples oyo epesami label.
Ndakisa: Kobongisa LLM mpo na kokoma ba e-mails ya mibeko nsima ya komonisa yango kaka ba exemples 10.
Fine-tuning (Fine-tuning)
Fine-tuning
Mokanda ya kozwa modèle oyo eyekolami liboso mpe koyekola yango lisusu na ensemble ya données ya sika, ya moke mpo na kosala ete ebonga mpo na mosala ya sikisiki.
Ndakisa: Kobongisa LLM ya général lokola GPT na ba documents ya mibeko ya kati mpo na kokela assistant ya kokoma mibeko.
Foundation Model (Foundation Model)
Foundation Model
Modèle ya monene oyo eyekolami na ba données ya ndenge na ndenge mpe ya monene oyo ekoki kobongisama na misala ebele ya nsima.
Ndakisa: GPT-4 mpe PaLM 2 ezali ba modèles ya fondation oyo ekoki kosala résumé, Q&A, traduction, mpe makambo mosusu.
Fuzzy Logic (Fuzzy Logic)
Fuzzy Logic
Lolenge ya logique oyo etali ba valeurs ya pene na esika ya logique ya solo/lokuta (binaire) ya sikisiki, ya motuya mpo na kokanisa na kozanga bosembo.
Ndakisa: Ekomami na ba systèmes ya contrôle ya climat mpo na kobongola température na kotalela ba entrées ya fuzzy lokola 'moke ya molunge' to 'malili mingi'.
Generative Adversarial Network (GAN) (Generative Adversarial Network (GAN))
Generative Adversarial Network (GAN)
Architecture ya modèle générative epai ba réseaux mibale — générateur mpe discriminateur — bazali kosala compétition mpo na kobongisa qualité ya sortie.
Ndakisa: Ba GANs esalelamaka mpo na kokela ba vidéos ya deepfake to kobimisa ba photos ya produit ya solo kobanda na ba croquis.
Generative AI (Generative AI)
Generative AI
Catégorie ya mayele ya kosala oyo ekoki kokela biloko ya sika — lokola texte, bililingi, miziki, to vidéo — kobanda na ba données ya koyekola.
Ndakisa: ChatGPT ezali kobimisa ba blog posts to Midjourney ezali kokela ba œuvres d'art numériques kobanda na ba prompts ya texte.
Generative Pre-trained Transformer (GPT) (Generative Pre-trained Transformer (GPT))
Generative Pre-trained Transformer (GPT)
Lolenge ya ba modèles ya lokota ya minene oyo ebimisami na OpenAI oyo esalelaka architecture ya transformer mpe eyekolami liboso na ba quantités ya minene ya ba données ya texte mpo na kosala misala ya lokota ya ndenge na ndenge.
Ndakisa: GPT-4 ekoki kokoma ba essais, kobongola ba lokota, mpe kosala résumé ya ba documents na prompt moke.
Genetic Algorithm (Genetic Algorithm)
Genetic Algorithm
Tekiniki ya optimisation oyo epesami bopusi na sélection naturelle epai ba solutions ebongwanaka na ntango na nzela ya mutation, crossover, mpe sélection.
Ndakisa: Ekomami mpo na kosala ba architectures ya réseau neuronal ya malamu na kosala simulation ya survie ya ba oyo bazali na makasi.
Hallucination (Hallucination)
Hallucination
Kobimisa biloko oyo ezali komonana lokola ya solo kasi ezali te ya solo to ya kozanga ntina na modèle ya AI.
Ndakisa: Modèle ya lokota ekelaka citation oyo ezali te to epesaka ba faits ya histoire ya lokuta.
Heuristic (Heuristic)
Heuristic
Nzela ya pratique ya kosala mikakatano oyo ekamwaka te solution ya kokoka kasi ekoki mpo na mikano ya ntango moke.
Ndakisa: Kosalela mibeko ya loboko mpo na kokanisa ntango ya livraison na système ya AI ya logistique.
Hyperparameter (Hyperparameter)
Hyperparameter
Valeur ya configuration oyo etiami liboso ya koyekola modèle ya koyekola ya masini, lokola taux ya koyekola to motango ya ba couches.
Ndakisa: Kobongola taille ya batch kobanda 32 tii 128 mpo na kobongisa vitesse ya koyekola mpe performance ya modèle.
Inference (Inference)
Inference
Mokanda ya kosalela modèle ya koyekola ya masini oyo eyekolami mpo na kosala ba prédictions to kobimisa ba résultats kobanda na ba données ya entrée ya sika.
Ndakisa: Kosalela modèle ya GPT oyo ebongisami mpo na kokoma ba e-mails mpo na équipe ya service client.
Intent Detection (Intent Detection)
Intent Detection
Mosala na compréhension ya lokota ya bomoto epai système eyebisaka mokano to ntina ya mokomi na message.
Ndakisa: Na chatbot, koyeba 'Nalingi kobuka vol' lokola mokano ya kobuka vol.
Internet of Things (IoT) (Internet of Things (IoT))
Internet of Things (IoT)
Réseau ya ba appareils physiques oyo esangisami na ba capteurs, logiciel, mpe ba tekiniki mosusu mpo na kosangisa mpe kobongola ba données.
Ndakisa: Ba thermostats ya mayele mpe ba frigos oyo ezali kopesa ba données ya kosalela mpe kobongola ba paramètres kosalelaka AI analytics.
Interpretability (Interpretability)
Interpretability
Niveau oyo moto akoki koyeba ba mécanismes ya kati ya modèle ya koyekola ya masini mpe processus na yango ya kozwa bikateli.
Ndakisa: Arbre ya décision ezali pɛtɛɛ koyeba koleka réseau neuronal ya mozindo mpo bikateli na yango ekoki kolandama.
Jupyter Notebook (Jupyter Notebook)
Jupyter Notebook
Environnement ya informatique interactive ya open-source oyo epesaka ba utilisateurs nzela ya kokoma code, kotala ba résultats, mpe kokoma analyse na interface moko.
Ndakisa: Ba scientifiques ya données basalelaka Jupyter Notebooks mpo na kosala ba prototypes ya ba modèles ya koyekola ya masini mpe kokabola ba résultats.
K-Nearest Neighbours (KNN) (K-Nearest Neighbours (KNN))
K-Nearest Neighbours (KNN)
Algorithme ya koyekola ya masini ya pɛtɛɛ, non-paramétrique oyo esalelamaka mpo na classification mpe régression. Ezalaka na bikateli na kotalela ba exemples ya koyekola ya pene na espace ya ba caractéristiques.
Ndakisa: Mpo na kokabola mbuma ya sika lokola pomme to poire, KNN etalaka ba mbuma nini oyo epesami label ezali pene na forme mpe langi.
Knowledge Graph (Knowledge Graph)
Knowledge Graph
Structure ya données oyo esalelaka ba nœuds mpe ba arêtes mpo na kopesa mpe kobomba ba descriptions ya ba entités oyo esangisami mpe ba relations na yango.
Ndakisa: Panneau ya connaissance ya Google epesami nguya na knowledge graph oyo esangisaka ba entités lokola bato, bisika, mpe ba événements.
Language Learning Model Optimisation (LLMO) (Language Learning Model Optimisation (LLMO))
Language Learning Model Optimisation (LLMO)
Ba tekiniki oyo esalelamaka mpo na kobongisa performance, efficacité, to adaptabilité ya ba modèles ya lokota ya minene mpo na misala to ba domaines ya sikisiki.
Ndakisa: Kosalela quantisation mpe instruction tuning mpo na kobongisa LLM mpo na kosalela na entreprise.
Large Language Model (LLM) (Large Language Model (LLM))
Large Language Model (LLM)
Lolenge ya modèle ya deep learning oyo eyekolami na ba quantités ya minene ya ba données ya texte oyo ekoki kobimisa, koyeba, mpe kokanisa na lokota ya bomoto.
Ndakisa: ChatGPT mpe Claude ezali ba LLM oyo eyekolami mpo na kosalisa na kokoma, coding, mpe koyanola mituna.
Latent Space (Latent Space)
Latent Space
Représentation abstraite ya dimension ya likolo epai ba entrées ya ndenge moko esangisami pene pene, esalelamaka na ba modèles génératifs mpe ba embeddings.
Ndakisa: Na génération ya bililingi, kobongola espace latent ekoki kobongola ba caractéristiques lokola luminosité to émotion.
Learning Rate (Learning Rate)
Learning Rate
Hyperparameter ya ntina na koyekola oyo ekontrolaka ndenge nini ba poids ya modèle ebongisami na kotalela gradient ya perte.
Ndakisa: Taux ya koyekola ya likolo ekoki komema na overshooting minima, nzokande taux ya moke mingi ekoki kokitisa progrès ya koyekola.
Machine Learning (ML) (Machine Learning (ML))
Machine Learning (ML)
Branche ya AI oyo epesaka ba systèmes nzela ya koyekola kobanda na ba données mpe kobongisa performance kozanga ete epesama programme ya polele.
Ndakisa: Ba filtres ya spam basalelaka koyekola ya masini mpo na kokabola ba e-mails lokola spam to te na kotalela ba exemples ya kala.
Model Drift (Model Drift)
Model Drift
Phénomène epai précision ya modèle ekitaka na ntango mpo na ba changements na ba données to environnement.
Ndakisa: Modèle ya koyeba fraude ekómaka moke ya précision lokola ba tactiques ya fraude ebongwanaka.
Model Training (Model Training)
Model Training
Mokanda ya kopesa ba données na modèle ya koyekola ya masini mpe kobongola ba paramètres na yango mpo na kokitisa erreur.
Ndakisa: Koyekola moteur ya recommandation na histoire ya kosomba ya ba clients mpo na kopesa ba produits ya sika.
Multimodal AI (Multimodal AI)
Multimodal AI
Ba systèmes ya AI oyo ekoki kosala mpe kosangisa ba types ya données ebele lokola texte, bililingi, audio, mpe vidéo.
Ndakisa: Modèle lokola GPT-4 Vision oyo ekoki kotanga texte mpe kotala bililingi na ntango moko.
Natural Language Processing (NLP) (Natural Language Processing (NLP))
Natural Language Processing (NLP)
Sous-domaine ya AI oyo etali interaction kati ya ba ordinateurs mpe ba lokota ya bomoto (naturel). Epesaka ba masini nzela ya kotanga, koyeba, mpe koyanola na lokota ya bomoto.
Ndakisa: NLP esalelamaka na ba assistants ya mongongo, ba applications ya traduction ya lokota, mpe ba chatbots.
Neural Network (Neural Network)
Neural Network
Modèle ya koyekola ya masini oyo epesami bopusi na structure ya bongo ya bomoto, esangisami na ba couches ya ba nœuds oyo esangisami (ba neurones).
Ndakisa: Ba réseaux neuronaux ezali na nsima ya ba modèles ya deep learning oyo esalelamaka na koyeba bililingi mpe mongongo.
Noise (Noise)
Noise
Ba informations ya hasard to ya kozanga ntina na ba données oyo ekoki kobombana ba modèles ya motuya mpe kobuka performance ya modèle.
Ndakisa: Ba erreurs ya capteur to ba entrées ya données oyo ezali na ba fautes ya kokoma ekoki kotalelama lokola bruit.
Ontology (Ontology)
Ontology
Cadre structuré oyo ekabolaka mpe elimbolaka ba relations kati ya ba concepts na domaine moko, mbala mingi esalelamaka na ba systèmes ya AI ya sémantique.
Ndakisa: Ontologie na santé ekoki kolimbola ndenge nini ba symptômes etali ba maladies mpe ba traitements.
Overfitting (Overfitting)
Overfitting
Erreur ya modélisation epai modèle ya koyekola ya masini ezwaka bruit na ba données ya koyekola mpe esalaka mabe na ba données ya sika.
Ndakisa: Modèle oyo ezali kokanisa ba eyano ya koyekola kasi ekoki te kosala na ba données ya test oyo emonani te ezali overfitted.
Predictive Analytics (Predictive Analytics)
Predictive Analytics
Kosalela ba données, ba algorithmes, mpe AI mpo na koyeba likoki ya ba résultats ya mikolo ezali koya na kotalela ba données ya kala.
Ndakisa: Ba détaillants basalelaka predictive analytics mpo na kokanisa demande ya ba produits ya sikisiki.
Pre-training (Pre-training)
Pre-training
Mokanda ya koyekola modèle liboso na ensemble ya données ya monene, ya général liboso ya kobongisa yango mpo na misala ya sikisiki.
Ndakisa: Ba modèles ya GPT eyekolami liboso na ba corpora ya minene liboso ya kobongisama mpo na ba chatbots ya service client.
Prompt Engineering (Prompt Engineering)
Prompt Engineering
Art mpe science ya kokela ba prompts ya malamu mpo na kolakisa sortie ya ba modèles ya lokota ya minene.
Ndakisa: Kobakisa ba instructions ya système lokola 'Yanolá lokola molakisi ya motema malamu' ezali ndakisa ya prompt engineering.
Quantisation (Quantisation)
Quantisation
Tekiniki ya compression ya modèle oyo ekitisaka motango ya ba bits oyo esalelamaka mpo na kopesa ba poids mpe ba activations, kobongisa efficacité.
Ndakisa: Kobongisa modèle kobanda 32-bit tii 8-bit ebongisaka performance na ba appareils mobiles.
Quantum Computing (Quantum Computing)
Quantum Computing
Paradigm ya sika ya informatique oyo etali mécanique quantique, oyo ezali na likoki ya ba capacités ya traitement ya exponential.
Ndakisa: Quantum computing ekoki mokolo moko kobongisa koyekola ya AI koleka ba limites ya classique.
Reasoning Engine (Reasoning Engine)
Reasoning Engine
Système na AI oyo ezwaka ba conclusions logiques kobanda na ensemble ya ba faits to ba données kosalelaka mibeko to ba algorithmes ya inférence.
Ndakisa: Outil ya diagnostic ya AI esalelaka reasoning engine mpo na kozwa ba conditions médicales oyo ekoki kozala na kotalela ba symptômes.
Reinforcement Learning (RL) (Reinforcement Learning (RL))
Reinforcement Learning (RL)
Domaine ya koyekola ya masini epai ba agents bayekolaka na kosala interaction na environnement na bango mpo na kobongisa ba récompenses ya cumulatif.
Ndakisa: Robot oyo ezali koyekola kotambola na nzela ya essai mpe erreur kosalelaka ba tekiniki ya RL.
Reinforcement Learning with Human Feedback (RLHF) (Reinforcement Learning with Human Feedback (RLHF))
Reinforcement Learning with Human Feedback (RLHF)
Méthode ya koyekola epai ba préférences ya bomoto ekambaka signal ya récompense ya AI, mbala mingi esalelamaka na fine-tuning ya ba modèles ya lokota.
Ndakisa: ChatGPT eyekolami na RLHF mpo na kobimisa ba eyano ya lisalisi mpe ya sécurité.
Retrieval-Augmented Generation (RAG) (Retrieval-Augmented Generation (RAG))
Retrieval-Augmented Generation (RAG)
Méthode oyo esangisaka information retrieval na génération, epai LLM ezwaka ba documents ya motuya mpo na kobongisa eyano na yango.
Ndakisa: Assistant ya AI azwaka mpe atangaka ba spécifications ya produit na ntango ya kobimisa eyano na motuna ya tekiniki.
Self-Supervised Learning (Self-Supervised Learning)
Self-Supervised Learning
Nzela ya koyekola epai modèle eyekolaka ba modèles na kobimisa ba labels na yango moko kobanda na ba données ya brut, kokitisa dépendance na ba données oyo epesami label na bomoto.
Ndakisa: BERT eyekolami na self-supervised learning na kokanisa maloba oyo ezangi na texte.
Semantic Search (Semantic Search)
Semantic Search
Tekiniki ya recherche oyo eyebaka mokano ya mokomi mpe motuya ya contexte, te kaka matching ya ba mots-clés.
Ndakisa: Koluka 'ndenge nini ya kobongisa robinet oyo ezali kobima mai' ezongisaka ba guides ata soki liloba 'robinet oyo ezali kobima mai' ezali te na document.
Sentiment Analysis (Sentiment Analysis)
Sentiment Analysis
Mokanda ya koyeba ba émotions, ba opinions, to ba attitudes na texte, mbala mingi kokabola lokola positif, négatif, to neutre.
Ndakisa: Kotala ba tweets mpo na koyeba réaction ya bato na produit ya sika.
Stochastic (Stochastic)
Stochastic
Ezali na hasard to bizaleli ya probabiliste, mbala mingi esalelamaka na AI générative mpe ba algorithmes ya optimisation.
Ndakisa: Sortie ya GPT-4 ebongwanaka mpo na entrée moko mpo na processus na yango ya décodage ya stochastic.
Strong AI (Strong AI)
Strong AI
Ekomami mpe lokola Artificial General Intelligence (AGI), etali ba masini oyo ezali na ba capacités cognitives ya niveau ya bomoto na ba domaines nyonso.
Ndakisa: AI ya mikolo ezali koya oyo ekoki kokoma ba romans na yango moko, kosala ba plans ya ba villes, mpe kosala mikakatano ya etika na ndenge moko.
Super Artificial Intelligence (SAI) (Super Artificial Intelligence (SAI))
Super Artificial Intelligence (SAI)
AI ya théorie oyo eleki mayele ya bomoto na makambo nyonso — kokanisa, créativité, mayele ya émotion, etc.
Ndakisa: SAI ekoki na théorie kokela ba sciences mpe ba philosophies ya sika na yango moko.
Supervised Learning (Supervised Learning)
Supervised Learning
Tekiniki ya koyekola ya masini epai ba modèles eyekolami na ba données oyo epesami label mpo na koyekola ba mappings ya entrée-sortie.
Ndakisa: Kolakisa modèle mpo na kokabola ba e-mails lokola spam to te kosalelaka ba exemples ya kala.
Synthetic Data (Synthetic Data)
Synthetic Data
Ba données oyo ebimisami na ndenge ya kosala simulation ya ba données ya solo, mbala mingi esalelamaka mpo na koyekola ntango ba données ya solo ezali moke to ya sensible.
Ndakisa: Kokela bililingi ya médical ya synthetic mpo na koyekola ba modèles ya diagnostic kozanga kobuka kobombama ya ba patients.
Token (Token)
Token
Unité ya texte oyo esalemi na ba LLM — mbala mingi liloba to eteni ya liloba.
Ndakisa: Phrase 'Hello world!' ekabwani na ba tokens 3: 'Hello', 'world', mpe '!'.
Tokenisation (Tokenisation)
Tokenisation
Mokanda ya kokabola texte na ba tokens mpo na kosala na modèle.
Ndakisa: Na NLP, 'ChatGPT is great' ekómaka ['Chat', 'G', 'PT', 'is', 'great'].
Transfer Learning (Transfer Learning)
Transfer Learning
Kosalela boyebi kobanda na mosala moko mpo na kobongisa koyekola na mosala mosusu oyo etali yango, kokitisa ntango ya koyekola mpe ba besoins ya données.
Ndakisa: Kobongisa modèle oyo eyekolami na texte ya Anglais mpo na kosala sentiment analysis na lokota mosusu.
Transformer (Transformer)
Transformer
Architecture ya réseau neuronal oyo esalelaka ba mécanismes ya attention mpo na kosala modèle ya ba données ya séquentiel, esalelamaka mingi na ba LLM.
Ndakisa: BERT, GPT, mpe T5 ezali ba modèles oyo etali transformer.
Underfitting (Underfitting)
Underfitting
Ntango modèle ezali pɛtɛɛ mingi mpo na kozwa ba modèles na ba données ya koyekola, kosalaka ete performance ezala mabe.
Ndakisa: Modèle ya linéaire oyo ezali komeka kokanisa ba classifications ya bililingi ya mikakatano ekoki kozala underfit.
Unsupervised Learning (Unsupervised Learning)
Unsupervised Learning
Nzela ya koyekola epai ba modèles eyebisaka ba modèles to ba clusters na ba données oyo epesami label te.
Ndakisa: Kokabola ba clients na kotalela bizaleli ya kosomba kozanga ba labels oyo elimbolami liboso.
User Intent (User Intent)
User Intent
Mokano to ntina ya motuna to interaction ya mokomi.
Ndakisa: Mokomi oyo akomi 'ndenge nini ya kosala gâteau' alingi kozwa recette.
Validation Set (Validation Set)
Validation Set
Sous-ensemble ya ba données oyo esalelamaka mpo na kotala performance ya modèle na ntango ya koyekola mpe kobongisa ba hyperparameters.
Ndakisa: Ekomami mpo na koyeba overfitting liboso ya test ya suka.
Vector Database (Vector Database)
Vector Database
Base ya données oyo esalemi mpo na kobomba mpe koluka ba embeddings ya vecteur oyo esalelamaka na ba misala ya AI lokola recherche ya similarité mpe RAG.
Ndakisa: Pinecone mpe Weaviate ezali ba bases de données ya vecteur mpo na kobomba ba embeddings ya texte to ya bililingi.
Vector Embedding (Vector Embedding)
Vector Embedding
Représentation numérique ya ba données oyo ebombaka motuya ya sémantique mpe ba relations na espace ya vecteur.
Ndakisa: Maloba 'mokonzi' mpe 'mwasi ya mokonzi' ezali na ba embeddings ya ndenge moko na ba différences ya moke ya lolenge ya mobali to ya mwasi.
Virtual Assistant (Virtual Assistant)
Virtual Assistant
Agent ya logiciel oyo esalaka na AI oyo esalisaka ba utilisateurs kosala misala na nzela ya kosolola to ba commandes ya mongongo.
Ndakisa: Siri, Alexa, mpe Google Assistant ezali ba assistants virtuels ya bato mingi.
Voice Recognition (Voice Recognition)
Voice Recognition
Tekiniki oyo etalaka mpe ebongolaka lokota elobami na texte to mosala.
Ndakisa: Kokoma na mongongo mpe ba commandes ya mongongo etali ba systèmes ya voice recognition.
Weak AI (Weak AI)
Weak AI
Ba systèmes ya AI oyo esalemi mpo na kosala mosala moko ya sikisiki kozanga mayele ya général.
Ndakisa: AI ya kosala échecs oyo ekoki te koyeba lokota to kolakisa motuka ezali ndakisa ya weak AI.
Web Scraping (Web Scraping)
Web Scraping
Extraction automatique ya ba informations kobanda na ba sites web, mbala mingi esalelamaka mpo na kosangisa ba données ya koyekola to kotala biloko.
Ndakisa: Kosala scraping ya ba annonces ya immobilier mpo na koyekola modèle ya évaluation ya propriété.
Weight (Weight)
Weight
Paramètre na ba réseaux neuronaux oyo elimbolaka makasi ya bopusi oyo nœud moko ezali na yango na mosusu.
Ndakisa: Ba poids ebongwanaka na ntango ya koyekola mpo na kokitisa erreur ya modèle.
Whisper (Whisper)
Whisper
Modèle ya mongongo na texte oyo ebimisami na OpenAI oyo ekoki kokoma audio na ba lokota ebele.
Ndakisa: Whisper ekoki kokoma ba conférences mpe ba podcasts na précision ya likolo.
YAML (YAML)
YAML
Format oyo moto akoki kotanga mpo na serialisation ya ba données, mbala mingi esalelamaka mpo na ba fichiers ya configuration na ba workflows ya koyekola ya masini.
Ndakisa: Kolimbola ba paramètres ya modèle na fichier YAML mpo na koyekola na PyTorch.
Zero-shot Learning (Zero-shot Learning)
Zero-shot Learning
Likoki ya modèle ya kosala misala oyo eyekolami te na polele na kosalela boyebi ya général.
Ndakisa: Modèle oyo ezali koyanola mituna ya mibeko ata soki eyekolami te na sikisiki na ba données ya mibeko.
Zettabyte (Zettabyte)
Zettabyte
Unité ya données numériques oyo ekokani na sextillion moko (10^21) ya ba octets, mbala mingi esalelamaka mpo na kolimbola taille ya ba données ya internet.
Ndakisa: Trafic ya internet ya mokili mobimba eleki 1 zettabyte na mobu na 2016.
