AI Words

Technology that lets computers do things that used to need a human mind.

Building AI by letting the computer learn rules from examples instead of being told.

Machine learning with neural networks stacked many layers deep.

Layers of simple multiply-and-add units wired together: the backbone of modern AI.

The trained core of an AI: a big bundle of numbers that turns an input into an answer.

Showing a model examples and nudging its numbers each time it gets one wrong.

Running a finished, trained model to get an answer out of it.

A collection of examples gathered to teach or test an AI.

One of the many numbers inside a model that get tuned during training.

A big AI model trained on huge amounts of text that writes by picking the next token.

The small piece of text an AI reads and writes, one at a time.

The tool that chops text into tokens and gives each piece an ID number.

Guessing the one token that comes next, based on all the text so far.

How much text an AI can read and keep in view at once, counted in tokens.

A setting that controls how adventurous an AI is when picking the next token.

Whatever you give an AI to work with: your question, instructions and examples.

When an AI confidently says something false or made up as if it were true.

The neural network design behind most of today's large language models.

The calculation that decides how much each token focuses on every other token.

A list of numbers that captures the meaning of a word, a sentence or an image.

The first, giant training stage where a model learns broadly from huge amounts of data.

Extra training that shapes an already-trained model for a specific job or style.

Training an AI on people's choices about which of two answers is better.

AI that creates new things: text, images, audio or video.

Describes an AI that handles more than text: images, sound and video too.

An AI trained to work through a problem step by step before it answers.

Getting an AI to write out its steps one by one before it gives the answer.

An AI calling outside tools, like a calculator, search or code, to get things right.

An AI that takes a goal and works through many steps on its own, using tools.

An open standard for plugging AI apps into tools and data in one shared way.

Having an AI look up relevant documents first, then answer based on them.

Hidden instructions an AI app gives the model before your conversation starts.

Building software by telling an AI what you want instead of writing the code yourself.

An AI that makes a picture by starting from random noise and cleaning it up step by step.

AI that listens to spoken words and turns them into written text.

The field of AI that lets computers look at images and work out what's in them.

A model whose trained numbers (weights) are published so anyone can download them.

A shared test that gives AI models the same questions so their scores can be compared.

The observation that AI gets predictably better as models, data and compute grow.

Training a small model to copy what a big model has learned.

Storing a model's numbers with less precision so it takes up less memory.

AI that runs right on your phone or laptop instead of on a faraway server.

A chip that does huge numbers of simple calculations at once: the workhorse of AI.

A building packed with computers where AI is trained and run.

The research and safeguards that aim to keep AI from causing harm.

Getting an AI to act the way people intend, in line with human values.

Wording a request cleverly to trick an AI into breaking its own safety rules.

Hiding instructions in content an AI reads, to make it do something it shouldn't.

A fake made with AI that convincingly copies a real person's face or voice.

When an AI's answers lean one way because the data it learned from was lopsided.

A hidden mark placed in AI-made content so it can later be recognised as AI-made.

AI that could handle almost any thinking task a person can, not just one or two.

The laws and rules governments set for how AI is built and used.

A test of whether you can tell a machine from a person just by chatting with them.

A 1958 artificial neuron that adjusts its own weights when it gets an example wrong.

Sending the error backward through a neural network to work out how to fix each weight.

A stretch of time when hopes for AI collapsed and funding and interest dried up.

A huge set of over 14 million labeled photos, and the yearly contest built around it.

Learning by trying things, getting rewards, and doing more of what earns them.

Machine learning from examples that come paired with the right answers.