4 ms·
Forgive my ignorance, but if the corpus of coding data was always 90% bad, isn't that the same data being used for training LLMs? How are they magically any bet
by lbrito 5mo ago
Forgive my ignorance, but if the corpus of coding data was always 90% bad, isn't that the same data being used for training LLMs? How are they magically any better than that average?
- epolanski 5mo agoBecause LLMs are not stochastic parrots.
- thoman23 5mo agoProgrammer: "What is this slop that I found in your code?" AI: "I LEARNED IT FROM YOU, DAD!"
- bigstrat2003 5mo agoThey aren't. The guy you're replying to is just hyping them up based on nothing.