[lbrito]

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?

[thoman23]

Programmer: "What is this slop that I found in your code?"

AI: "I LEARNED IT FROM YOU, DAD!"

[bigstrat2003]

They aren't. The guy you're replying to is just hyping them up based on nothing.

[epolanski]

Because LLMs are not stochastic parrots.