AI systems capable of autonomous action—perceiving their environment, making decisions, and executing multi-step tasks with minimal human intervention.

Legal frameworks, policies, and standards governing the development and deployment of artificial intelligence systems.

Research and practices aimed at ensuring AI systems behave as intended and don't cause unintended harm, from near-term risks to existential concerns.

The practice of making AI systems understandable—disclosing how they work, what data they use, and how decisions are made.

Techniques to embed invisible markers in AI-generated content to enable identification of its synthetic origin.

Systematic errors in AI systems that create unfair outcomes, typically reflecting biases present in training data or design choices.

Application Programming Interface—a set of rules allowing software systems to communicate, enabling developers to access AI capabilities programmatically.

Software that simulates conversation with users, ranging from rule-based scripts to sophisticated AI assistants powered by LLMs.

The computational resources—processing power, memory, and infrastructure—required to train and run AI models.

Technologies and standards for verifying the origin, history, and integrity of digital content in an era of AI-generated media.

The practice of monitoring and filtering user-generated content to enforce platform policies and legal requirements.

The maximum amount of text an LLM can process at once—the model's working memory that limits how much it can 'see' in a conversation.

Legal protection granting creators exclusive rights to their original works, now central to debates over AI training data and outputs.

Deceptive user interface designs that manipulate users into actions they didn't intend, from unwanted purchases to privacy violations.

The right of individuals to control how their personal information is collected, used, and shared by organizations and AI systems.

Synthetic media created using AI to realistically depict someone saying or doing something they never actually did.

A generative AI architecture that creates images by learning to reverse a gradual noising process, powering systems like Stable Diffusion and DALL-E.

False or misleading information deliberately created and spread to deceive, distinct from misinformation which may be shared unknowingly.

A numerical representation of text, images, or other data as vectors, enabling AI to measure similarity and meaning mathematically.

A security method where only the communicating parties can read messages, with no access possible for service providers or intermediaries.

The gradual degradation of online platforms as they shift value from users to advertisers and shareholders, ultimately harming everyone.

AI technology that identifies or verifies individuals by analyzing facial features, raising significant privacy and civil liberties concerns.

The process of further training a pre-trained AI model on domain-specific data to specialize its capabilities for particular tasks.

AI systems that create new content—text, images, audio, video, or code—rather than just analyzing or classifying existing data.

When an AI model generates false or fabricated information presented as fact, a key reliability challenge in language models.

The process of running a trained AI model to generate predictions or outputs from new inputs—the 'using' phase as opposed to training.

An AI system trained on vast text datasets to understand and generate human-like text, powering applications from chatbots to code generation.

A subset of AI where systems learn patterns from data rather than being explicitly programmed with rules.

False or inaccurate information spread without intent to deceive, distinct from disinformation which is deliberately misleading.

The challenge of ensuring AI systems behave according to human values and intentions, avoiding harmful or unintended behaviors.

AI systems that can process and generate multiple types of content—text, images, audio, and video—within a single model.

A computing system inspired by biological brains, using layers of interconnected nodes to learn patterns from data.

AI models and systems released with publicly available weights, code, and documentation, allowing inspection, modification, and redistribution.

The practice of crafting effective instructions for AI models to elicit desired outputs, a key skill for working with language models.

Retrieval-Augmented Generation—a technique that enhances LLM responses by first retrieving relevant information from external sources.

AI systems that predict what content, products, or connections users will find relevant, powering feeds across social media, streaming, and e-commerce.

Reinforcement Learning from Human Feedback—a technique to align AI behavior with human preferences by training on human evaluations.

The integration of social media with television viewing, enabling real-time audience engagement and second-screen experiences.

An economic system where personal data is extracted and commodified for behavioral prediction and manipulation, often without meaningful consent.

Any media—images, video, audio, or text—generated or substantially modified by AI rather than captured or created by humans.

AI systems that generate images from written descriptions, transforming text prompts into visual content.

AI systems that generate video content from text descriptions, representing a frontier in generative AI.

The dataset used to teach an AI model patterns and behaviors—the examples from which it learns.

A neural network architecture using self-attention mechanisms, forming the foundation of modern LLMs like GPT and Claude.

AI technology that replicates a person's voice from audio samples, enabling synthetic speech that sounds like the original speaker.