Comment by ggus
15 days ago
It's very hard to have ChatGPT et al tell me that an idea I had isn't good.
I have to tailor my prompts to curb the bias, adding a strong sense of doubt on my every idea, to see if the thing stops being so condescending.
15 days ago
It's very hard to have ChatGPT et al tell me that an idea I had isn't good.
I have to tailor my prompts to curb the bias, adding a strong sense of doubt on my every idea, to see if the thing stops being so condescending.
Maybe "idea evaluation" is just a bad use case for LLMs?
Most times the idea is implied. I'm trying to solve a problem with some tools, and there are better tools or even better approaches.
ChatGPT (and copilot and gemini) instead all tell me "Love the intent here — this will definitely help. Let's flesh out your implementation"...
Qualitative judgment in general is probably not a great thing to request from LLMs. They don't really have a concept of "better" or "worse" or the means to evaluate alternate solutions to a problem.