If you’ve been following the series, here are 5 more terms. If not, here’s 5 more anyway.
Token
- A unit of input for an AI model
- Kinda like a unit of money is a penny (in the US at least) or a unit of language is a letter
Tokenization
- The process that AI tools use to break down input into tokens
- For example, if the input was the word “asparagus,” it might be tokenized as ‘aspar’ and ‘agus’
- There are actually a lot of different ways that input can be tokenized (by letter, by word, by subword, etc.)
- The model used determines how things are tokenized
Context Window
- The number of tokens that a model can process at one time
- If you go over this number, the model will start to “forget” information
Inference
- An AI tool guessing what you want
- Similar to human inference. Think about any time someone says, “You know what I mean”
API (Application Programming Interface)
- A way for applications to talk to each other
*Note - You could argue that this is still kinda techie. You’d probably be right. On the bright side, imagine how complicated the full techie definitions might be
Comments