Skip to main content
Zubnet AI › Learn › Wiki

AI Wiki — Page 3 of 10

AI concepts explained by builders, not textbooks. No jargon walls. No academic gatekeeping. Just clear, practical definitions of the terms you'll actually encounter.

454 terms 8 categories Updated 2026-11-22
💡
Term of the Day
Deepgram
Speech AI company building fast, accurate speech recognition and text-to-speech APIs. Their Nova models compete with and often beat OpenAI's Whisper…
→
No terms on this page match your filters.

101–150 of 454 terms · Page 3 of 10

Filter this page

Companies
A specialized cloud provider built entirely around GPU computing for AI workloads. CoreWeave operates large clusters…
CorpusAlso known as: Dataset, Training Data
Training
The body of text (or other data) used to train a model. A corpus can range from curated collections of books and…
CorrigibilityAlso known as: Shutdown Problem, Off-Switch Problem
Safety
The property of an AI system that accepts correction, modification and shutdown by its legitimate overseers instead of…
Cosine SimilarityCosine Distance, Vector Similarity
Fundamentals
A measure of similarity between two vectors based on the angle between them, ignoring their magnitude. Cosine…
Cross-AttentionEncoder-Decoder Attention
Fundamentals
An attention mechanism where the queries come from one sequence and the keys/values come from a different sequence. In…
Cross-EntropyAlso known as: Cross-Entropy Loss, Log Loss, Negative Log-Likelihood
Fundamentals
The loss function nearly every language model is trained on, which penalizes the model according to how little…
Cross-ValidationK-Fold CV, Leave-One-Out
Training
A technique for evaluating model performance when you don't have enough data for a separate test set. K-fold…
CUDAAlso known as: Compute Unified Device Architecture
Infrastructure
NVIDIA's platform for running general-purpose code on its GPUs, and the software foundation that most AI runs on.
A training strategy that presents examples in a meaningful order — typically from easy to hard — rather than randomly.…
Companies
An AI-native code editor built as a fork of VS Code, designed around the idea that code editing should be a…
DALL-EDALL-E 2, DALL-E 3
Models
OpenAI's image generation model family. DALL-E 1 (2021) used a discrete VAE + Transformer approach. DALL-E 2 (2022)…
Techniques that artificially expand a training dataset by creating modified versions of existing examples. For images:…
Data CentersAlso known as: AI Data Centers, GPU Clusters
Infrastructure
Physical facilities that house the servers, GPUs, networking equipment, and cooling systems needed to train and run AI…
Data PoisoningAlso known as: Training Data Poisoning, Backdoor Attack
Safety
An attack that slips malicious examples into a model's training data to plant hidden behaviors or degrade its…
DatabricksMosaic ML, DBRX, Unity Catalog
Companies
A data and AI platform that provides unified analytics, data engineering, and machine learning capabilities.…
DatasetTraining Set, Data
Fundamentals
A structured collection of data used to train, evaluate, or test a machine learning model. Datasets can be labeled…
Decart AIAlso known as: Real-time world simulation, game generation
Companies
Israeli AI company pushing the boundaries of real-time AI generation. Their technology can generate interactive…
DecoderDecoder Network, Generator
Fundamentals
A neural network component that generates output from a representation. In Transformers, the decoder uses causal…
Deep LearningAlso known as: DL
Fundamentals
A subset of machine learning that uses neural networks with many layers (hence "deep") to learn hierarchical…
Deep ResearchAlso known as: Deep Research Agents, Agentic Research
Using AI
A mode offered by several AI assistants where the model autonomously plans and carries out a multi-step web research…
DeepfakesAlso known as: Synthetic Media, AI-Generated Fakes
Safety
AI-generated images, video, or audio designed to convincingly depict real people saying or doing things they never…
DeepgramAlso known as: Nova speech-to-text, Aura text-to-speech
Companies
Speech AI company building fast, accurate speech recognition and text-to-speech APIs. Their Nova models compete with…
DeepLAlso known as: Neural machine translation, DeepL Pro
Companies
German AI company widely regarded as the best machine translation service in the world. Built by a team of…
DeepSeekAlso known as: DeepSeek-V3, DeepSeek-R1
Companies
Chinese AI lab that shook the industry in early 2025 with DeepSeek-R1, a reasoning model rivaling frontier labs at a…
Developer ToolsAlso known as: AI SDKs, AI Frameworks
Tools
The ecosystem of libraries, frameworks, and platforms that make building AI-powered applications easier. This includes…
DevinAlso known as: Cognition Devin, Cognition AI
Tools
An autonomous AI software engineer built by Cognition AI that takes high-level tasks and carries them out end to end:…
A mathematical framework that guarantees individual privacy in aggregate data analysis and model training. With…
Diffusion LLMAlso known as: Text Diffusion Model, dLLM
Models
A large language model that generates text through iterative denoising instead of predicting one token at a time from…
A type of generative model that creates images (or video, audio) by starting with pure noise and gradually removing it…
An architecture that replaces the U-Net backbone traditionally used in diffusion models with a Transformer. DiT…
DistillationKnowledge Distillation, Model Distillation
Training
Training a smaller "student" model to mimic the behavior of a larger "teacher" model. Instead of training the student…
Distributed TrainingData Parallelism, Model Parallelism, FSDP
Infrastructure
Training a model across multiple GPUs or machines simultaneously. Data parallelism gives each GPU a copy of the model…
DPODirect Preference Optimization
Training
An alternative to RLHF for aligning language models with human preferences. Instead of training a separate reward…
Drift DetectionData Drift, Model Drift, Concept Drift
Infrastructure
Monitoring for changes in the data distribution or model behavior over time that could degrade performance. Data…
DropoutWeight Decay
Training
A regularization technique that randomly "turns off" a fraction of neurons during each training step by setting their…
Dual UseDual-Use Technology
Safety
Technology that can be used for both beneficial and harmful purposes. AI is inherently dual-use: the same model that…
Early StoppingPatience, Validation-Based Stopping
Training
Stopping training when performance on a held-out validation set stops improving, rather than training for a fixed…
Edge AIOn-Device AI, Local AI
Infrastructure
Running AI models directly on end-user devices — phones, laptops, IoT sensors, cars — rather than in the cloud. Edge…
ElevenLabsAlso known as: Voice synthesis, voice cloning, dubbing
Companies
Voice AI company that made ultra-realistic speech synthesis accessible to everyone. Their technology powers voice…
EmbeddingAlso known as: Vector Embedding
Training
A way to represent text (or images, or audio) as a list of numbers (a vector) that captures its meaning. Similar…
Embedding LayerToken Embedding, Embedding Table, Lookup Table
Fundamentals
A lookup table that maps each token in the vocabulary to a dense vector (the token's embedding). When the model…
Embodied AIAlso known as: Embodied Intelligence, Robotics Foundation Models
Fundamentals
An approach to artificial intelligence where models control physical machines — robots that perceive, move, and…
EmergenceAlso known as: Emergent Abilities, Emergent Behavior
Fundamentals
Capabilities that appear in AI models at scale but were not explicitly trained for — abilities that seem to "emerge"…
EncoderEncoder Network, Feature Extractor
Fundamentals
A neural network component that converts input data into a compressed, information-rich representation (encoding). In…
Encoder-DecoderSeq2Seq, Sequence-to-Sequence
Models
A model architecture with two distinct parts: an encoder that reads and compresses the input into a representation,…
Infrastructure
A specific URL where an AI API accepts requests. For example, Anthropic's message endpoint is where you send prompts…
EU AI ActAlso known as: AI Act, European AI Act
Safety
The European Union's comprehensive law regulating artificial intelligence, in force since August 2024. It takes a…
EvaluationAlso known as: Evals, Model Evaluation
Training
The methods used to measure how well an AI model performs. This goes far beyond benchmarks — it includes human…
Existential RiskX-Risk, AI Doom
Safety
The hypothesis that sufficiently advanced AI systems could pose a threat to human existence or permanently curtail…
ESC