How LLMs Actually Work
Intro · 4 concepts · 4 clips · 6 source videos. Watch the clips in order, then take the quick check to see what did not land.
Tokens: what the model actually reads
A model never sees words. It sees tokens, which is why it miscounts letters, why context has a size, and why usage is billed the way it is.
Matt Pocock takes tokenisation from characters to subwords, and shows where unusual words go wrong (2:16-9:45)
Did this clip teach it?