Damus
Open in Damus
dbat :godot: profile picture
dbat :godot:
@dbat :godot:
Question. If llm training on images requires (abused, poor) humans to identify/describe/tag every image, how does llm training on code work?

How does concept map onto code?
1
Critter (he/him) · 4w
nostr:nprofile1qy2hwumn8ghj7un9d3shjtnyd968gmewwp6kyqpq7yr7pmc0a3h7a0ykx5d04j0zdc7js0mksy8ht7wvehwpht6wh83q3d2uyx I am hoping it's based on reading the comments so we can start poisoning the well by writing bad code with comments that explain the code solving the wrong problems and push these absolu...