FrontierBasics
Learn the ideas behind frontier technology, one short lesson at a time. Each track starts from the basics and builds up. Free for everyone.
Generative AI
- 1 AI model A program that has learned patterns from examples, so it can make predictions on new input. Beginner
- 2 training Teaching a model by showing it many examples and correcting its mistakes. Beginner
- 3 token A small piece of text, often part of a word, that language models read and write. Beginner
- 4 large language model A very large AI model trained on huge amounts of text to understand and write language. Beginner
- 5 prompt The instructions and text you give an AI model to get the answer you want. Beginner
- 6 context window The maximum amount of text a model can read at once before answering. Beginner
- 7 embedding A list of numbers that captures the meaning of a piece of text, so similar ideas sit close together. Intermediate
- 8 autoregressive Writing an answer one piece at a time, each piece chosen based on what came before. Intermediate
- 9 non-autoregressive Producing an entire answer in one step, not word by word. Intermediate
- 10 forward pass One trip of an input through a trained model to get its answer. Intermediate
- 11 hallucination When an AI model states something false with full confidence. Intermediate
- 12 fine-tuning Extra training that adapts an existing model to one specific job, using your own examples. Intermediate
- 13 retrieval-augmented generation Letting a model look up relevant documents first, then answer using them. Advanced
AI Operations
- 1 inference Using a trained model to get an answer, as opposed to training it. Beginner
- 2 latency How long you wait between sending a request and getting the answer back. Beginner
- 3 GPU A computer chip built to do many calculations at once, used to train and run AI models. Beginner
- 4 model API A way for your software to send text to an AI model hosted by someone else and get an answer back. Beginner
- 5 checkpoint A saved, trained version of a model, ready to run without retraining. Beginner
- 6 open-weight model A model whose learned numbers are published, so anyone can download and run it themselves. Beginner
- 7 batching Processing several requests together in one go, to use the hardware more efficiently. Intermediate
- 8 router Software that automatically sends each request to the right one of several trained models. Intermediate
- 9 quantization Shrinking a model by storing its numbers with less precision, so it runs faster and cheaper. Intermediate
- 10 benchmark A standard test used to measure and compare how well models perform. Intermediate