Glossary

A branch of AI where computers learn patterns from data rather than being explicitly programmed with rules. Powers most modern AI applications.

A type of AI model trained on vast amounts of text to predict the next word in a sequence. Powers ChatGPT, Claude, Gemini, and similar tools.

A computational architecture loosely inspired by the brain, made of layered artificial neurons that learn patterns from data. The foundation of modern AI.

Hypothetical AI that can match or exceed human cognitive abilities across essentially all tasks, rather than being narrowly specialised.

A technique combining language models with external retrieval. Lets AI systems look up specific documents at query time rather than relying solely on training data.

A neural network architecture introduced in 2017 that uses self-attention. The foundation of all modern large language models.

A neural network trained on huge amounts of text that can generate, summarize, translate, and reason about language. Examples: GPT-5, Claude, Gemini, Llama.

The neural network architecture behind every modern AI model. Uses 'attention' to process sequences in parallel. Introduced in 2017 by Google researchers.

A technique that grounds LLM responses in external documents by retrieving relevant context before generation. Reduces hallucinations.

A numerical representation of text, images, or other data as a vector. Similar items have similar vectors — the foundation of semantic search and RAG.

The process of continuing the training of a pre-trained model on specific data to specialize it for a task or domain.

When an AI model generates plausible-sounding but factually incorrect information. A fundamental limitation of LLMs.

A training technique that uses human ratings of model outputs to make AI more helpful, harmless, and honest. Critical to making ChatGPT and Claude useful.

The problem of ensuring AI systems pursue goals that match human values and intentions. Central concern in AI safety research.

The practice of crafting effective inputs to LLMs to get desired outputs. Combines clear instructions, context, examples, and format specifications.

A neural network architecture where multiple specialized 'expert' sub-networks exist, with a router selecting which to use for each input. Enables massive models at lower compute cost.

A database optimized for storing and searching high-dimensional vectors (embeddings). Powers semantic search, RAG, and recommendations.

An AI system that takes multi-step actions to accomplish goals — browsing the web, running code, editing files — rather than just responding to single prompts.

A type of generative AI that creates images (and now video) by progressively denoising random noise. Powers DALL-E, Stable Diffusion, Midjourney, Sora.

The process of breaking text into smaller units (tokens) that LLMs process. Tokens are typically sub-word fragments, not whole words.

The maximum amount of text an LLM can consider at once. Measured in tokens. Larger windows enable longer documents and conversations.

Deoxyribonucleic acid — the molecule that stores genetic instructions for all known life. Made of four chemical bases (A, T, G, C) arranged in a double helix.

A precise gene-editing technology derived from a bacterial immune system. Allows scientists to edit DNA sequences in living cells with letter-level accuracy.

Ribonucleic acid — a single-stranded molecule that carries genetic instructions from DNA to the cell's protein-making machinery, and serves many other regulatory roles.

A medical technique that treats disease by altering or replacing a patient's genes. Approved therapies now exist for several previously fatal genetic disorders.

A vaccine that delivers synthetic messenger RNA into cells, instructing them to temporarily produce a viral protein that trains the immune system.

The process by which populations of organisms change over generations, primarily through natural selection acting on inherited variation.

The collective community of microorganisms — bacteria, archaea, fungi, viruses — living on and in a host organism.

Interest earned on both the original principal and previously accumulated interest. The mathematical engine behind long-term wealth and unmanageable debt.

The rate at which prices rise over time, reducing money's purchasing power. Mild inflation is normal; high inflation is destructive.

A sustained decline in economic activity, conventionally defined as two consecutive quarters of negative GDP growth.

Units of ownership in a corporation, traded on exchanges. Long-term, broad stock markets have historically been the highest-returning major asset class.

Debt securities — loans made to governments or companies that pay periodic interest and return the principal at maturity.

The first decentralised cryptocurrency, secured by cryptography and consensus rules. Operates without any central authority since 2009.

The branch of physics describing the behaviour of matter and energy at the smallest scales — atoms, electrons, photons. Predicts probabilities rather than certainties.

Einstein's 1915 theory of gravity, describing it as the curvature of spacetime caused by mass and energy. Predicts black holes, gravitational waves, and time dilation.

A measure of the number of microscopic arrangements consistent with a system's macroscopic state. The second law of thermodynamics says it always increases over time.

The process of combining light atomic nuclei into heavier ones, releasing enormous energy. Powers stars; pursued for decades as a clean energy source on Earth.

A region of spacetime where gravity is so strong that nothing — not even light — can escape. Formed when massive stars collapse at the end of their lives.

A planet that orbits a star other than our Sun. Over 5,500 confirmed as of 2025, with thousands more candidates awaiting verification.

The catastrophic explosion of a star at the end of its life, briefly outshining an entire galaxy and forging most heavy elements in the universe.

An incredibly dense stellar remnant left after a core-collapse supernova. About 20 km wide but containing more mass than the Sun.

The leading scientific model for the origin of the universe — an extraordinary expansion of space itself starting ~13.8 billion years ago.

An invisible form of matter making up about 27% of the universe. Detected only through its gravitational effects on galaxies and large-scale structure.

A mysterious force or property of space causing the universe's expansion to accelerate. Makes up about 68% of total mass-energy.

Matter composed of antiparticles — for every regular particle, there is a mirror partner with opposite charge. Annihilates with normal matter on contact.

Long-term shifts in Earth's average temperature and weather patterns, currently driven primarily by human emissions of greenhouse gases.

Technology that removes CO₂ from emission sources or directly from the air, storing it underground or using it as feedstock.

Energy sources that replenish naturally — solar, wind, hydro, geothermal, biomass — and produce far fewer emissions than fossil fuels.

A distributed, append-only database where each new block of records cryptographically links to the previous one. Powers cryptocurrencies like Bitcoin.

A computer that uses quantum-mechanical phenomena (superposition, entanglement) to perform certain calculations exponentially faster than classical computers.