3 ms·
That is no different from pretty any other person in the world. If I interview people to catch them on mistakes, I will be able to do exactly that. Sure, there
by wickedsight 1y ago
That is no different from pretty any other person in the world. If I interview people to catch them on mistakes, I will be able to do exactly that. Sure, there are some exceptions, like if you were to interview Linus about Linux. Other than that, you'll always be able to find a fluke in someone's knowledge.
None of this makes me 'snap out' of anything. Accepting that LLM's aren't perfect means you can just keep that in mind. For me, they're still a knowledge multiplier and they allow me to be more productive in many areas of life.
- tecleandor 1y agoNot at all. Useful or not, LLMs will almost never say "I don't know". They'll happily call a function to a library that never existed. They'll tell you "Incredible idea! You're on the correct path! And you can easily do that with so and so software", and you'll be like "wait what, that software doesn't do that", and they'll answer "Ah, yeah, you're right, of course."
- yujzgzc 1y agoNo there are many techniques now to curb hallucinations. Not perfect but no longer so egregiously overconfident.
- ninkendo 1y ago…such as?
- ignoramous 1y agoTFA says, hallucinations is why "gyms" will be important: Language tooling (compiler, linter, language server, domain-specific static analyses etc) that feed back into the Agent, so it'll know to redo.
- rvnx 1y agoSometimes asking in a loop: "are you sure ? think step-by-step", "are you sure ? think step-by-step", "are you sure ? think step-by-step", "are you sure ? think step-by-step", "verify the result" or similar, you may end up with "I'm sure yes", and then you know you have a quality answer.
- rvnx 1y agoThe most infuriating are the emojis everywhere