U vhambedza (Alignment)
Alignment
Ndzila ya u vhona uri sisiteme ya AI i vhambedzana na zwipikwa zwa muthu, mvelelo, na vhuḓifari. Hezwi zwi vhukuma zwa ndeme kha sisiteme dzo bveledzwaho dzi no nga dza bveledza vhuḓifari vhu sa ṱoḓiwi nga ho livhaho.
Tsumbo: U vhona uri chatbot ya thuso ya mutakalo wa muhumbulo a i ngo vhuya ya khou á¹±uá¹±uwedza zwiito zwi vhavhaho naho zwi tshi khou á¹±uá¹±uwedzwa.
Application Programming Interface (API) (Application Programming Interface (API))
Application Programming Interface (API)
Muvhigo wa milayo yo á¹±alutshedzwaho na milayo ine ya tendela sisiteme dzo fhambanaho dza software u amba na u shandulana data.
Tsumbo: U shumisa OpenAI API u ruma prompt na u wana pfunzo ya luambo yo bveledzwaho nga mushini kha app yawe ya web.
Artificial General Intelligence (AGI) (Artificial General Intelligence (AGI))
Artificial General Intelligence (AGI)
Fomo ya AI ya theoretical ine ya kona u ita mushumo naho u ufhio wa vhuá¹±ali une muthu a kona u u ita. I generalises pfunzo kha domains.
Tsumbo: Sisiteme ya AGI i nga guda u vhumba muzika, u ita mushumo wa u vhulahisa, na u fhira exam ya philosophy hu si na programming yo livhaho mushumo.
Artificial Intelligence (AI) (Artificial Intelligence (AI))
Artificial Intelligence (AI)
U lingedza vhuṱali ha muthu kha mimushini yo programwaho u humbula, u humbula, na u ita nga u ḓiimisela.
Tsumbo: AI i shuma kha vhathusi vha muthu vha ngaho Siri na sisiteme dza u ḓiṱwaḓisa dza ngaho Tesla Autopilot.
AI Ethics (AI Ethics)
AI Ethics
Pfunzo i vhambedzanaho na zwiá¹±havhuso zwa vhumorali zwa u bveledza na u shumisa AI, u katela u luga, privacy, u vhona, na u sa khethulula.
Tsumbo: U sika milayo ya u thivhela hiring algorithms u khethulula nga ha dzhene kana vhukale.
Augmented Intelligence (Augmented Intelligence)
Augmented Intelligence
Muvhigo wa u shumisana hune AI ya thusa na u khwinisa vhuá¹±ali ha muthu ná¹±hani ha u vhu dzhia.
Tsumbo: AI-powered radiology tools dzi no sumbedza anomalies kha madokotela, vhane vha ita diagnosis ya u fhedza.
Autonomous Agent (Autonomous Agent)
Autonomous Agent
Sisiteme ya AI i konaho u ḓiitela zwiṱhavhuso na u dzhia zwiito u swikelela zwipikwa zwayo hu si na u dzhenelela ha muthu.
Tsumbo: Robotho ya u ḓiṱwaḓisa i no ṱwaḓisa zwitarata zwa ḓorobo na u thivhela zwivhili nga u ḓiimisela.
Backpropagation (Backpropagation)
Backpropagation
Ndzila ya u guda mitambo ya neural nga u shandula weights nga u vhambedza u bva kha output u swika kha input layers, u fhungudza vhukhakhi ha u vhona.
Tsumbo: I shumiswa kha u guda image classifiers u fhungudza error rate kha u vhona nomboro dzo ṅwalwaho nga tshanḓa.
Bias (Algorithmic Bias) (Bias (Algorithmic Bias))
Bias (Algorithmic Bias)
U khethulula hu si na u ṱoḓa na hu systematic kha mvelelo dza AI nga nṱhani ha data ya u guda yo sa vhambedzanaho kana yo sa imelaho.
Tsumbo: Sisiteme ya u vhona tshifhaá¹±uwo i no khakhela vhathu vha muvhala kanzhi nga ná¹±hani ha u sa imelwa kha data ya u guda.
Big Data (Big Data)
Big Data
Data yo vuleaho vhukuma ine ya ṱoḓa zwishumiswa zwo livhaho u vhulunga, u ṱolisisa, na u bvisa vhuimo, kanzhi i shumiswa u guda AI models.
Tsumbo: U shumisa dzimilioni dza u shumisana ha vhashumisi u guda recommendation engines dza e-commerce platforms.
Black Box Model (Black Box Model)
Black Box Model
Muvhigo wa AI kana pfunzo ya mushini une logic yawo ya nga ngomu i sa pfesesei nga vhathu, zwi tshi ita uri zwi konḓe u pfesesa uri zwiṱhavhuso zwi itwa hani.
Tsumbo: Neural network yo dzikaho i shumiswa u tendela loans fhedzi i sa ṋei ṱhaluso yo livhaho ya uri ngani muṱoḓi o tendelwa na muṅwe o landulwa.
Cognitive Computing (Cognitive Computing)
Cognitive Computing
Sisiteme dza AI dzo itelwaho u lingedza ndila dza u humbula dza muthu, dza ngaho u humbula na u guda, nga u shumisa ndila dza ngaho NLP na pattern recognition.
Tsumbo: Sisiteme ya cognitive computing i no thusa vha-legal professionals u á¹±olisisa case law na u vhona mvelelo.
Computer Vision (Computer Vision)
Computer Vision
Pfunzo ya vhuá¹±ali ha vhutsila ine ya tendela khomphuyutha u á¹±alutshedza na u shandula data ya u vhona yo ngaho zwifanyiso na video.
Tsumbo: Sisiteme dza u vhona tshifhaá¹±uwo dzi no vhona vhathu kha security footage nga u shumisa computer vision.
Corpus (Corpus)
Corpus
Tsumbo khulu ya maá¹…walwa kana maipfi o ambiwaho a shumiswaho u guda language models.
Tsumbo: Common Crawl dataset ndi public web corpus i shumiswaho u guda large language models dza ngaho GPT.
Data Drift (Data Drift)
Data Drift
Tshiitiko hune input data ya shanduka nga tshifhinga, zwi tshi ita uri model performance i fhungudzwe.
Tsumbo: Muvhigo wa predictive maintenance wa industrial equipment u vha u sa luga nga u fhelela musi hu tshi dzheniswa thekinolodzhi ntswa ya sensor.
Data Labelling (Data Labelling)
Data Labelling
Ndzila ya u á¹…wala data nga tags kana labels u itela uri i kone u shumiswa kha supervised learning.
Tsumbo: U á¹…wala zwikete zwa zwifanyiso zwa tumour sa benign kana malignant u guda cancer detection model.
Data Mining (Data Mining)
Data Mining
Ndzila ya u wana patterns dza ndeme, correlations, na anomalies kha large datasets.
Tsumbo: Vharengisi vha shumisa data mining u vhona uri vhathu vhane vha renga nappies kanzhi vha renga na beer.
Deep Learning (Deep Learning)
Deep Learning
Pfunzo ya mushini i shumisaho multi-layered neural networks u vhumba patterns dzo tsikidzanaho kha data.
Tsumbo: Deep learning i shumiswa kha language models dza ngaho GPT-4 na image generation models dza ngaho Stable Diffusion.
Diffusion Models (Diffusion Models)
Diffusion Models
Muvhigo wa generative models dzi no guda u bveledza data nga u shandula noise ya random kha structured outputs.
Tsumbo: Stable Diffusion i sika zwifanyiso zwa photorealistic u bva kha text prompts nga u shumisa diffusion techniques.
Embedding (Embedding)
Embedding
Muvhigo wa numerical vector wa data, kanzhi u shumiswa u dzhia semantic meaning ya maipfi, zwifanyiso, kana sentense.
Tsumbo: Kha NLP, ipfi 'bank' li nga vha na embeddings dzi fanaho na 'money' fhedzi dzo fhambanaho na 'riverbank' nga u vhambedza na context.
Epoch (Epoch)
Epoch
U vhambedza ho fhelelaho kha training dataset yo fhelelaho nga tshifhinga tsha training process ya machine learning model.
Tsumbo: Arali dataset i na 1,000 examples na model i vhona dzoá¹±he lwa u thoma nga tshifhinga tsha training, yeneyo ndi epoch nthihi.
Ethical AI (Ethical AI)
Ethical AI
Design na deployment philosophy ine ya vhona uri thekinolodzhi dza AI dzi shuma nga u vhonala, nga u luga, na nga u vhambedzana na values dza tshitshavha.
Tsumbo: AI hiring tool i no katela bias checks u thivhela u khethulula vhaṱoḓi vha minority.
Expert System (Expert System)
Expert System
Sisiteme ya AI i no lingedza vhukoni ha u ita zwiá¹±havhuso ha human expert kha specific domain nga u shumisa milayo na logic.
Tsumbo: Expert system i shumiswa kha vhulimi u á¹±uá¹±uwedza crop treatments nga u vhambedza na soil data na pest history.
Explainable AI (XAI) (Explainable AI (XAI))
Explainable AI (XAI)
Sisiteme dza AI dzo itelwaho u ita uri internal processes na zwiá¹±havhuso zwa dzo zwi pfesesei kha vhathu, zwi tshi engedza trust na accountability.
Tsumbo: AI ya medical diagnostic i no ṋea recommendation fhedzi i tshi ṱalutshedza uri ndi zwifhio zwiṱhavhuso zwe zwa ita uri hu vhe na yeneyo mvelelo.
Few-shot Learning (Few-shot Learning)
Few-shot Learning
Ndzila ya pfunzo ya mushini hune model ya gudiwa kana ya fine-tuned nga u shumisa nomboro á¹±hukhu fhedzi ya labelled examples.
Tsumbo: U shandula LLM u á¹…wala emails dza legal nga murahu ha u i sumbedza fhedzi 10 examples.
Fine-tuning (Fine-tuning)
Fine-tuning
Ndzila ya u dzhia pre-trained model na u i guda nga vhuya kha new, smaller dataset u i specialise kha specific task.
Tsumbo: U fine-tuning general LLM ya ngaho GPT kha internal legal documents u sika legal drafting assistant.
Foundation Model (Foundation Model)
Foundation Model
Muvhigo muhulu wo gudiwaho kha data yo fhambanaho na yo vuleaho ine ya nga shandulwa kha mishumo minzhi ya downstream.
Tsumbo: GPT-4 na PaLM 2 ndi foundation models dzi konaho u summarisation, Q&A, translation, na zwiá¹…we.
Fuzzy Logic (Fuzzy Logic)
Fuzzy Logic
Fomo ya logic i no vhambedzana na approximate values ná¹±hani ha fixed true/false (binary) logic, i shumiseaho kha u humbula hu si na vhukuma.
Tsumbo: I shumiswa kha climate control systems u shandula temperature nga u vhambedza na fuzzy inputs dza ngaho 'a bit hot' kana 'very cold'.
Generative Adversarial Network (GAN) (Generative Adversarial Network (GAN))
Generative Adversarial Network (GAN)
Muvhigo wa generative model hune mitambo mivhili — generator na discriminator — ya vhambedzana u khwinisa output quality.
Tsumbo: GANs dzi shumiswa u sika deepfake videos kana u bveledza realistic product photos u bva kha sketches.
Generative AI (Generative AI)
Generative AI
Kategori ya vhuṱali ha vhutsila ine ya kona u sika content ntswa — yo ngaho text, zwifanyiso, muzika, kana video — u bva kha training data.
Tsumbo: ChatGPT i no bveledza blog posts kana Midjourney i no sika digital artwork u bva kha textual prompts.
Generative Pre-trained Transformer (GPT) (Generative Pre-trained Transformer (GPT))
Generative Pre-trained Transformer (GPT)
Kategori ya large language models dzo bveledzwaho nga OpenAI dzi no shumisa transformer architecture na dzo pre-trained kha vhunzhi ha text data u ita mishumo yo fhambanaho ya luambo.
Tsumbo: GPT-4 i kona u á¹…wala essays, u shandula luambo, na u summarise documents nga minimal prompting.
Genetic Algorithm (Genetic Algorithm)
Genetic Algorithm
Ndzila ya optimisation yo á¹±uá¹±uwedzwaho nga natural selection hune solutions dza bveledza nga tshifhinga nga u shanduka, crossover, na selection.
Tsumbo: I shumiswa u design efficient neural network architectures nga u lingedza survival of the fittest.
Hallucination (Hallucination)
Hallucination
U bveledza ha plausible-sounding fhedzi factually incorrect kana nonsensical content nga AI model.
Tsumbo: Language model i no sika non-existent citation kana u ṋea false historical facts.
Heuristic (Heuristic)
Heuristic
Ndzila ya practical ya u tandulula thaidzo ine ya sa vhoni solution yo fhelelaho fhedzi i no fanelela zwipikwa zwa zwino.
Tsumbo: U shumisa rule of thumb u vhona delivery time kha logistics AI system.
Hyperparameter (Hyperparameter)
Hyperparameter
Value ya configuration yo vheiwaho phanḓa ha u guda machine learning model, yo ngaho learning rate kana nomboro ya layers.
Tsumbo: U shandula batch size u bva kha 32 u swika kha 128 u khwinisa training speed na model performance.
Inference (Inference)
Inference
Ndzila ya u shumisa trained machine learning model u ita predictions kana u bveledza outputs u bva kha new input data.
Tsumbo: U shumisa fine-tuned GPT model u draft emails dza customer support team.
Intent Detection (Intent Detection)
Intent Detection
Mushumo kha natural language understanding hune sisiteme ya vhona goal kana purpose ya mushumisi kha message.
Tsumbo: Kha chatbot, u vhona 'I want to book a flight' sa travel booking intent.
Internet of Things (IoT) (Internet of Things (IoT))
Internet of Things (IoT)
Network ya interconnected physical devices dzo dzheniswaho na sensors, software, na thekinolodzhi dziá¹…we u vhulunga na u shandulana data.
Tsumbo: Smart thermostats na fridges dzi no vhiga usage data na u shandula settings nga u shumisa AI analytics.
Interpretability (Interpretability)
Interpretability
U vhambedza hune muthu a kona u pfesesa internal mechanics ya machine learning model na decision-making process yawo.
Tsumbo: Decision tree i pfesesea u fhira deep neural network ngauri zwiá¹±havhuso zwayo zwi nga vhonwa.
Jupyter Notebook (Jupyter Notebook)
Jupyter Notebook
Open-source interactive computing environment i no tendela vhashumisi u á¹…wala code, u vhona outputs, na u á¹…wala analysis kha single interface.
Tsumbo: Data scientists vha shumisa Jupyter Notebooks u prototype machine learning models na u share results.
K-Nearest Neighbours (KNN) (K-Nearest Neighbours (KNN))
K-Nearest Neighbours (KNN)
Simple, non-parametric machine learning algorithm i shumiswaho kha classification na regression. I ita zwiá¹±havhuso nga u vhambedza na closest training examples kha feature space.
Tsumbo: U classify new fruit sa apple kana pear, KNN i vhona uri ndi zwifhio zwifhuga zwo labelled zwi re tsini kha shape na colour.
Knowledge Graph (Knowledge Graph)
Knowledge Graph
Data structure i no shumisa nodes na edges u imela na u vhulunga interlinked descriptions dza entities na relationships dzadzo.
Tsumbo: Google's knowledge panel i powered nga knowledge graph i no vhambedza entities dza ngaho vhathu, fhethu, na zwiitiko.
Language Learning Model Optimisation (LLMO) (Language Learning Model Optimisation (LLMO))
Language Learning Model Optimisation (LLMO)
Ndzila dzi shumiswaho u khwinisa performance, efficiency, kana adaptability ya large language models kha specific tasks kana domains.
Tsumbo: U shumisa quantisation na instruction tuning u optimise LLM for enterprise use.
Large Language Model (LLM) (Large Language Model (LLM))
Large Language Model (LLM)
Muvhigo wa deep learning model wo gudiwaho kha vhunzhi ha textual data i konaho u bveledza, u pfesesa, na u humbula nga luambo lwa muthu.
Tsumbo: ChatGPT na Claude ndi LLMs dzo gudiwaho u thusa kha u á¹…wala, u coding, na u fhindula mbudziso.
Latent Space (Latent Space)
Latent Space
High-dimensional abstract representation hune similar inputs dza vhulungwa tsini, dzi shumiswa kha generative models na embeddings.
Tsumbo: Kha image generation, u shandula latent space zwi nga shandula features dza ngaho brightness kana emotion.
Learning Rate (Learning Rate)
Learning Rate
Key hyperparameter kha training i no vhona uri model weights dzi shandulwa hani nga u vhambedza na loss gradient.
Tsumbo: High learning rate i nga ita uri hu vhe na overshooting minima, ngeno low rate i tshi slow training progress.
Machine Learning (ML) (Machine Learning (ML))
Machine Learning (ML)
Pfunzo ya AI ine ya tendela sisiteme u guda u bva kha data na u khwinisa performance hu si na u programwa nga ho livhaho.
Tsumbo: Spam filters dzi shumisa machine learning u classify emails sa spam kana not nga u vhambedza na past examples.
Model Drift (Model Drift)
Model Drift
Tshiitiko hune accuracy ya model ya fhungudzea nga tshifhinga nga ná¹±hani ha u shanduka ha data kana environment.
Tsumbo: Fraud detection model i vha i sa luga nga u fhelela musi fraud tactics dzi tshi bveledza.
Model Training (Model Training)
Model Training
Ndzila ya u ṋea data kha machine learning model na u shandula parameters dzayo u fhungudza error.
Tsumbo: U guda recommendation engine kha customer purchase history u á¹±uá¹±uwedza new products.
Multimodal AI (Multimodal AI)
Multimodal AI
Sisiteme dza AI dzi konaho u shandula na u dzhenisa multiple types of data yo ngaho text, zwifanyiso, audio, na video.
Tsumbo: Muvhigo wa ngaho GPT-4 Vision une wa kona u vhala text na u á¹±alutshedza zwifanyiso nga tshifhinga tshi fanaho.
Natural Language Processing (NLP) (Natural Language Processing (NLP))
Natural Language Processing (NLP)
Pfunzo ya AI i livhaho kha u shumisana ha khomphuyutha na luambo lwa muthu (natural). I tendela mimushini u vhala, u pfesesa, na u fhindula nga luambo lwa muthu.
Tsumbo: NLP i shumiswa kha voice assistants, language translation apps, na chatbots.
Neural Network (Neural Network)
Neural Network
Machine learning model yo á¹±uá¹±uwedzwaho nga structure ya vhongo ya muthu, yo vhumbwaho nga layers dza interconnected nodes (neurons).
Tsumbo: Neural networks dzi vhukuma ha deep learning models dzi shumiswaho kha image na speech recognition.
Noise (Noise)
Noise
Mafhungo a random kana a sa vhambedzanaho kha data ane a nga vhona patterns dza ndeme na u vhukuma u vhukuma model performance.
Tsumbo: Sensor errors kana typo-filled data entries zwi nga vhonwa sa noise.
Ontology (Ontology)
Ontology
Structured framework i no categorise na u á¹±alutshedza relationships kha mihumbulo kha domain, kanzhi i shumiswa kha semantic AI systems.
Tsumbo: Ontology kha healthcare i nga á¹±alutshedza uri symptoms dzi vhambedzana hani na diseases na treatments.
Overfitting (Overfitting)
Overfitting
Modelling error hune machine learning model ya dzhia noise kha training data na u shuma nga u sa luga kha new data.
Tsumbo: Muvhigo une wa memorise training answers fhedzi u sa kone u handle unseen test data ndi overfitted.
Predictive Analytics (Predictive Analytics)
Predictive Analytics
U shumisa data, algorithms, na AI u vhona likelihood ya future outcomes nga u vhambedza na historical data.
Tsumbo: Vharengisi vha shumisa predictive analytics u forecast demand for certain products.
Pre-training (Pre-training)
Pre-training
Ndzila ya u guda model kha large, general dataset phanḓa ha u i fine-tuning kha specific tasks.
Tsumbo: GPT models dzi pre-trained kha large corpora phanḓa ha u customised for customer service chatbots.
Prompt Engineering (Prompt Engineering)
Prompt Engineering
Vhutsila na science ya u sika effective prompts u steer output ya large language models.
Tsumbo: U engedza system instructions dza ngaho 'Reply as a polite tutor' ndi tsumbo ya prompt engineering.
Quantisation (Quantisation)
Quantisation
Model compression technique i no fhungudza nomboro ya bits dzi shumiswaho u imela weights na activations, i tshi engedza efficiency.
Tsumbo: U quantising model u bva kha 32-bit u swika kha 8-bit zwi khwinisa performance kha mobile devices.
Quantum Computing (Quantum Computing)
Quantum Computing
New paradigm ya computing nga u vhambedza na quantum mechanics, ine ya vha na potential ya exponential processing capabilities.
Tsumbo: Quantum computing i nga ḽiṅwe ḓuvha ya accelerate AI training u fhira classical limits.
Reasoning Engine (Reasoning Engine)
Reasoning Engine
Sisiteme kha AI ine ya bvisa logical conclusions u bva kha set of facts kana data nga u shumisa milayo kana inference algorithms.
Tsumbo: AI diagnosis tool i shumisa reasoning engine u deduce possible medical conditions nga u vhambedza na symptoms.
Reinforcement Learning (RL) (Reinforcement Learning (RL))
Reinforcement Learning (RL)
Pfunzo ya mushini hune agents dza guda nga u shumisana na environment yadzo u maximise cumulative rewards.
Tsumbo: Robotho i no guda u ṱwaḓisa nga trial and error nga u shumisa RL techniques.
Reinforcement Learning with Human Feedback (RLHF) (Reinforcement Learning with Human Feedback (RLHF))
Reinforcement Learning with Human Feedback (RLHF)
Ndzila ya pfunzo hune human preferences dza vhona AI's reward signal, kanzhi i shumiswa kha fine-tuning language models.
Tsumbo: ChatGPT yo gudiwa nga RLHF u bveledza helpful na safe responses.
Retrieval-Augmented Generation (RAG) (Retrieval-Augmented Generation (RAG))
Retrieval-Augmented Generation (RAG)
Ndzila i no vhambedza information retrieval na generation, hune LLM ya toda relevant documents u khwinisa response yayo.
Tsumbo: AI assistant i toda na u cita product specs musi i tshi bveledza answer kha technical question.
Self-Supervised Learning (Self-Supervised Learning)
Self-Supervised Learning
Ndzila ya training hune model ya guda patterns nga u bveledza labels dzayo u bva kha raw data, i tshi fhungudza reliance kha human-annotated data.
Tsumbo: BERT yo gudiwa nga self-supervised learning nga u vhona missing words kha text.
Semantic Search (Semantic Search)
Semantic Search
Search technique i no pfesesa user intent na contextual meaning, hu si keyword matching fhedzi.
Tsumbo: U toda 'how to fix a leaking tap' zwi vhiga guides naho ipfi 'leaking tap' li siho kha document.
Sentiment Analysis (Sentiment Analysis)
Sentiment Analysis
Ndzila ya u vhona emotions, opinions, kana attitudes kha text, kanzhi i classify sa positive, negative, kana neutral.
Tsumbo: U á¹±olisisa tweets u vhona public reaction kha new product.
Stochastic (Stochastic)
Stochastic
U katela randomness kana probabilistic behaviour, kanzhi i shumiswa kha generative AI na optimisation algorithms.
Tsumbo: Output ya GPT-4 i fhambana kha input i fanaho nga ná¹±hani ha stochastic decoding process yayo.
Strong AI (Strong AI)
Strong AI
I dovha ya vhidzwa Artificial General Intelligence (AGI), i vhambedzana na mimushini i re na human-level cognitive capabilities kha domains dzoá¹±he.
Tsumbo: AI ya tshifhinga tshi ḓaho i no kona u ḓiṅwalela novels, u plan cities, na u tandulula ethical dilemmas nga u luga.
Super Artificial Intelligence (SAI) (Super Artificial Intelligence (SAI))
Super Artificial Intelligence (SAI)
AI ya theoretical i no fhira vhuṱali ha muthu kha zwiṅwe zwiṅwe zwoṱhe — u humbula, u sika, emotional intelligence, etc.
Tsumbo: SAI i nga theoretically ya bveledza new sciences na philosophies nga u ḓiimisela.
Supervised Learning (Supervised Learning)
Supervised Learning
Machine learning technique hune models dza gudiwa kha labelled data u guda input-output mappings.
Tsumbo: U funza model u classify emails sa spam kana not nga u shumisa historical examples.
Synthetic Data (Synthetic Data)
Synthetic Data
Artificially generated data i no lingedza real-world data, kanzhi i shumiswa kha training musi real data i tshi vha scarce kana sensitive.
Tsumbo: U sika synthetic medical images u guda diagnostic models hu si na u vhukuma patient privacy.
Token (Token)
Token
Unit ya text yo shandulwaho nga LLMs—kanzhi ipfi kana word piece.
Tsumbo: Sentense 'Hello world!' i fhandekana kha 3 tokens: 'Hello', 'world', na '!'.
Tokenisation (Tokenisation)
Tokenisation
Ndzila ya u fhandekanya text kha tokens u itela u shandulwa nga model.
Tsumbo: Kha NLP, 'ChatGPT is great' i vha ['Chat', 'G', 'PT', 'is', 'great'].
Transfer Learning (Transfer Learning)
Transfer Learning
U shumisa nḓivho u bva kha mushumo muthihi u khwinisa pfunzo kha muṅwe mushumo wo vhambedzanaho, u fhungudza training time na data needs.
Tsumbo: Fine-tuning model yo gudiwaho kha English text u ita sentiment analysis kha muá¹…we luambo.
Transformer (Transformer)
Transformer
Neural network architecture i no shumisa attention mechanisms u vhumba sequential data, i shumiswa nga vhunzhi kha LLMs.
Tsumbo: BERT, GPT, na T5 ndi models dzo vhumbwaho nga transformer.
Underfitting (Underfitting)
Underfitting
Musi model i tshi vha yo leluwaho vhukuma u dzhia patterns kha training data, zwi tshi ita uri hu vhe na poor performance.
Tsumbo: Linear model i no lingedza u vhona complex image classifications i nga underfit.
Unsupervised Learning (Unsupervised Learning)
Unsupervised Learning
Ndzila ya pfunzo hune models dza vhona patterns kana clusters kha unlabelled data.
Tsumbo: U vhulunga vhashumisi nga u vhambedza na purchasing behaviour hu si na predefined labels.
User Intent (User Intent)
User Intent
Goal kana purpose ya user's query kana interaction.
Tsumbo: Mushumisi a no ṅwala 'how to bake a cake' zwi nga vha zwi tshi khou ṱoḓa recipe.
Validation Set (Validation Set)
Validation Set
Subset ya data i shumiswaho u evaluate model performance nga tshifhinga tsha training na tune hyperparameters.
Tsumbo: I shumiswa u vhona overfitting phanḓa ha final testing.
Vector Database (Vector Database)
Vector Database
Database yo itelwaho u vhulunga na u toda vector embeddings dzi shumiswaho kha AI tasks dza ngaho similarity search na RAG.
Tsumbo: Pinecone na Weaviate ndi vector databases dza u vhulunga text kana image embeddings.
Vector Embedding (Vector Embedding)
Vector Embedding
Numeric representation ya data i no vhulunga semantic meaning na relationships kha vector space.
Tsumbo: Maipfi 'king' na 'queen' a na embeddings dzi fanaho na subtle gender differences.
Virtual Assistant (Virtual Assistant)
Virtual Assistant
AI-powered software agent i no thusa vhashumisi u fhedza mishumo nga u amba kana voice commands.
Tsumbo: Siri, Alexa, na Google Assistant ndi virtual assistants dzi no ḓivhea.
Voice Recognition (Voice Recognition)
Voice Recognition
Thekinolodzhi i no á¹±alutshedza na u shandula luambo lwo ambiwaho kha text kana action.
Tsumbo: Voice typing na voice commands zwi vhambedzana na voice recognition systems.
Weak AI (Weak AI)
Weak AI
Sisiteme dza AI dzo itelwaho u ita narrow, specific task hu si na general intelligence.
Tsumbo: Chess-playing AI i no kona u pfesesa luambo kana u ṱwaḓisa goloi ndi tsumbo ya weak AI.
Web Scraping (Web Scraping)
Web Scraping
Automated extraction ya information u bva kha websites, kanzhi i shumiswa u vhulunga training data kana u monitor content.
Tsumbo: Scraping real estate listings u guda property valuation model.
Weight (Weight)
Weight
Parameter kha neural networks i no vhona uri strength ya influence ya node nthihi i na kha iá¹…we.
Tsumbo: Weights dzi shanduka nga tshifhinga tsha training u fhungudza model's error.
Whisper (Whisper)
Whisper
Speech-to-text model yo bveledzwaho nga OpenAI i konaho u transcribe audio kha multiple languages.
Tsumbo: Whisper i kona u transcribe lectures na podcasts nga high accuracy.
YAML (YAML)
YAML
Human-readable format ya data serialisation, kanzhi i shumiswa kha configuration files kha machine learning workflows.
Tsumbo: U á¹±alutshedza model parameters kha YAML file ya training kha PyTorch.
Zero-shot Learning (Zero-shot Learning)
Zero-shot Learning
Vhukoni ha model u ita tasks dze ya sa vhuya ya gudiwa nga ho livhaho nga u shumisa general knowledge.
Tsumbo: Muvhigo u no fhindula legal questions naho u sa gudiwa nga ho livhaho kha legal data.
Zettabyte (Zettabyte)
Zettabyte
Unit ya digital data i no vhambedzana na one sextillion (10^21) bytes, kanzhi i shumiswa u á¹±alutshedza scale ya internet data.
Tsumbo: Global internet traffic yo fhira 1 zettabyte nga á¹…waha nga 2016.
