Well I mean it’s that or find more guys willing to go through it manually, which seems to be the problem since it’s open source. Unless they can scrounge up the money to hire people to do it full time.
It’s frequently hard to tell at a glance codegen slop, you actually have to look at it and understand what’s going on. An LLM that would produce such slop itself isn’t going to be effective at detecting such slop.
Sounds like they need a bot to check the code for AI telltales. Send AI to kill AI.
Sounds like an excellent use of power and water and cou cycles in data centers.
Well I mean it’s that or find more guys willing to go through it manually, which seems to be the problem since it’s open source. Unless they can scrounge up the money to hire people to do it full time.
It’s frequently hard to tell at a glance codegen slop, you actually have to look at it and understand what’s going on. An LLM that would produce such slop itself isn’t going to be effective at detecting such slop.