- DepartureAI basics
- Stop 2Learning from examples
- Stop 3Prompts
- Stop 4Tokens
- Stop 5Embeddings
- Stop 6Attention
- Stop 7Next-token probabilities
- Stop 8Repeated generation
- Stop 9Temperature
- ArrivalChecking answers
Platform 01 · Station 5 of 10
Embeddings
An embedding is a list of numbers that helps a computer compare meanings. Think of a map where related ideas are close together. The map is only an example, not a real place.
Find a passage with a similar meaning
Question: Where can I borrow a book? Its example embedding is [0.8, 0.2]. The passage Borrow books from our library has a similar example embedding [0.7, 0.3], so it is a better match. The passage Football practice at 4 pm has a different example embedding [0.1, 0.9]. These number lists are invented for teaching, not calculated by a real model.
02 · Try
Which passage is most related to “Where can I borrow a book?”?
Progress saves in this browser.