AI Glossary

An AI system that can take actions autonomously to achieve goals, rather than just responding to prompts.

The study of moral questions raised by AI development and deployment, including fairness, accountability, transparency, and societal impact.

The field focused on ensuring AI systems work as intended without causing unintended harm to humans or society.

The challenge of making AI systems behave in ways that match human values and intentions.

AI technique that identifies unusual patterns or outliers that don't conform to expected behavior.

Anthropic's interface for accessing Claude models, offering AI assistance with a focus on safety and helpfulness.

A way for software programs to communicate with each other, allowing developers to integrate AI capabilities into their applications.

A technique that allows neural networks to focus on the most relevant parts of the input when producing each part of the output.

Microsoft's framework for building multi-agent systems where AI agents collaborate through conversation.

AI systems that can operate independently to achieve goals with minimal human oversight or intervention.

Microsoft's enterprise service providing access to OpenAI models through Azure with added security and compliance features.

An algorithm that calculates how much each weight in a neural network contributed to errors, enabling the network to learn from mistakes.

AWS's managed service for accessing foundation models from multiple providers through a unified API.

Bidirectional Encoder Representations from Transformers, Google's model designed for understanding text context by reading in both directions simultaneously.

Systematic patterns in AI outputs that unfairly favor or disadvantage particular groups, often reflecting biases present in training data or design choices.

A software application that simulates human conversation through text or voice interactions.

DeepMind's language model that proved smaller, better-trained models outperform larger undertrained ones, reshaping how the field thinks about scaling.

Anthropic's family of AI assistants designed with constitutional AI principles, emphasizing helpfulness, harmlessness, and honesty.

An AI feature that predicts and suggests code as developers type, speeding up programming workflows.

OpenAI's code-specialized model that powered GitHub Copilot, trained to understand and generate programming code across dozens of languages.

An enterprise-focused AI company providing language models, embeddings, and retrieval systems optimized for business applications.

Cohere's retrieval-augmented generation optimized model designed specifically for enterprise RAG applications and tool use.

A field of AI that enables computers to interpret and understand visual information from images and videos.

A training approach where AI systems are given explicit principles to follow and learn to critique and revise their own outputs accordingly.

The maximum amount of text an AI model can consider at once, including both your input and its response.

A framework for orchestrating autonomous AI agents that work together as a crew to accomplish complex tasks.

NVIDIA's parallel computing platform that enables developers to use GPUs for general-purpose processing and AI workloads.

OpenAI's text-to-image model that generates original images from natural language descriptions with strong prompt adherence.

Techniques for artificially expanding training datasets by creating modified versions of existing data, like rotating images or paraphrasing text.

A subset of machine learning that uses neural networks with many layers to learn complex patterns from large amounts of data.

A Chinese AI lab's open-source language models known for strong coding abilities and competitive performance at efficient training costs.

A generative AI approach that creates images by learning to gradually remove noise from random static until a coherent image emerges.

A training technique where a smaller student model learns to mimic the behavior of a larger teacher model.

AI technology that extracts, classifies, and structures information from documents like PDFs, forms, and images.

Numerical representations of text that capture semantic meaning, allowing AI to find similar content and understand relationships.

The ability to understand and communicate why an AI system made a particular decision or produced a specific output.

The Technology Innovation Institute's open-source language models from Abu Dhabi, known for high-quality pretraining data and competitive benchmarks.

The ability of an AI model to learn new tasks or concepts from just a handful of examples, rather than requiring thousands of training samples.

The process of training an existing AI model on specific data to customize its behavior for particular tasks.

Google DeepMind's multimodal AI model family designed to natively understand and generate text, images, audio, and video.

AI systems that create new content like text, images, audio, or video based on patterns learned from training data.

Generative Pre-trained Transformer, OpenAI's family of large language models that generate human-like text through autoregressive prediction.

A Graphics Processing Unit that accelerates AI model training and inference through parallel computation.

An optimization algorithm that iteratively adjusts model parameters to minimize errors by moving in the direction of steepest improvement.

Connecting AI responses to authoritative sources or real-world data to reduce hallucinations and improve accuracy.

When an AI generates information that sounds plausible but is factually incorrect or made up.

A platform and company providing open-source ML libraries, pre-trained models, and infrastructure for AI development.

AI technique that assigns labels or categories to images based on their visual content.

The process of using a trained model to make predictions or generate outputs on new, previously unseen data.

A deployed API endpoint that serves a trained AI model for real-time predictions.

The degree to which humans can understand the internal workings and reasoning processes of an AI model.

A prompt or technique designed to bypass an AI system's safety guidelines and get it to produce normally restricted content.

A framework for building applications powered by language models, with tools for chaining prompts, memory, and external data.

A compressed, abstract representation of data where similar items are positioned near each other and meaningful operations become possible.

Large Language Model Meta AI, Meta's open-weight language model family that enabled widespread research and commercial fine-tuning of powerful AI systems.

A data framework for connecting custom data sources to large language models through indexing and retrieval.

A type of AI trained on massive amounts of text that can understand and generate human language.

A branch of artificial intelligence where computers learn patterns from data to make predictions or decisions without being explicitly programmed.

AI technology that automatically translates text or speech from one language to another.

A server that implements the Model Context Protocol, allowing AI models to interact with external tools and data sources in a standardized way.

A proprietary text-to-image AI service known for producing highly aesthetic, artistic images through a Discord-based interface.

A French AI company's efficient open-weight language models known for punching above their weight class in performance per parameter.

Mistral AI's Mixture of Experts model that achieves GPT-3.5 level performance while using a fraction of the compute per inference.

The process of deploying and managing machine learning models to handle prediction requests in production.

AI systems that can understand and generate multiple types of data, such as text, images, audio, and video, within a single model.

AI technique that identifies and classifies named entities like people, organizations, and locations in text.

A field of AI focused on enabling computers to understand, interpret, and generate human language.

A computing system inspired by the human brain that learns patterns from data through interconnected nodes organized in layers.

AI technique that identifies and locates multiple objects within an image or video, drawing bounding boxes around each.

Optical Character Recognition technology that converts images of text into machine-readable characters.

Open Neural Network Exchange, an open format for representing machine learning models across different frameworks.

OpenAI's interface for accessing GPT models, DALL-E, Whisper, and other AI capabilities programmatically.

When a model learns training data too perfectly, including noise and quirks, causing poor performance on new unseen data.

Pathways Language Model, Google's large-scale language model that preceded Gemini and demonstrated strong reasoning and multilingual capabilities.

Microsoft's family of small language models that achieve surprisingly strong performance through high-quality synthetic training data.

Using statistical models and machine learning to forecast future outcomes based on historical data.

The practice of crafting inputs to AI models to get better, more consistent, and more useful outputs.

An attack where malicious instructions are hidden in input data, tricking an AI system into ignoring its original instructions.

An optimization technique that removes unnecessary weights or neurons from a neural network to reduce size and computation.

A technique that reduces model size and speeds up inference by using lower-precision numbers for weights and activations.

AI system that automatically answers questions posed in natural language based on given context or knowledge.

Alibaba Cloud's family of multilingual language models with strong performance in Chinese and English, released as open weights.

A technique that enhances AI responses by retrieving relevant information from a knowledge base before generating answers.

AI system that predicts and suggests items or content a user might like based on their behavior and preferences.

The practice of deliberately trying to find flaws, vulnerabilities, and harmful behaviors in AI systems before they're deployed.

A machine learning approach where agents learn optimal behavior through trial and error, receiving rewards or penalties for their actions.

Reinforcement Learning from Human Feedback, a training technique where AI learns to improve its responses based on human ratings and preferences.

Microsoft's open-source SDK for integrating large language models into applications with plugins and planners.

AI technique that classifies every pixel in an image into a category, creating detailed visual maps.

AI technique that identifies and categorizes emotions and opinions expressed in text.

Technology that converts spoken audio into written text, also known as automatic speech recognition.

An open-source text-to-image diffusion model that generates detailed images from text descriptions and runs on consumer hardware.

AI technique that condenses long documents or text into shorter versions while preserving key information.

A machine learning approach where models learn from labeled examples that include both inputs and their correct outputs.

Artificially generated data that mimics the statistical properties of real data, used to train AI models when actual data is scarce or sensitive.

AI technology that generates images from written text descriptions.

Technology that converts written text into spoken audio using synthesized or cloned voices.

AI technique that predicts future values based on patterns observed in sequential historical data.

The basic units that AI models use to process text, roughly corresponding to word fragments or characters.

The ability of an AI model to call external functions, APIs, or services to accomplish tasks beyond text generation.

Google's custom Tensor Processing Unit, an ASIC designed specifically for accelerating machine learning workloads.

The dataset used to teach a machine learning model patterns and relationships, forming the foundation of what the model learns.

A neural network architecture that processes sequences using self-attention, enabling models to weigh the importance of different parts of the input.

When a model is too simple to capture the underlying patterns in the data, resulting in poor performance on both training and new data.

A machine learning approach where models discover hidden patterns and structures in data without labeled examples.

A database optimized for storing and searching embeddings, enabling fast similarity search across millions of items.

Google Cloud's unified machine learning platform for building, deploying, and scaling AI models.

An AI-powered software agent that performs tasks and services based on voice or text commands.

OpenAI's open-source automatic speech recognition model that transcribes and translates audio with near-human accuracy across multiple languages.

The ability of an AI model to perform tasks it was never explicitly trained on, using only instructions or descriptions without any examples.