Hacker News Reader

Comment by hristov | Hacker News Reader

HN should have a [chat GPT] tag for stuff generated by chat GPT. So that I can ignore it. I do not mind reading what paul graham said about how to get new ideas. I do not care what chat GPT said. What chat GPT said may or may not be a good summary of paul grahams thoughts, it may be wrong in an important aspect, etc.

If Paul Graham wants to say something, he should say it directly and not hide behind a chat gpt layer.

View on HN

This is PG saying something... he linked to a chatGPT generated response and said he didn't like the response so he wrote his own.

PG was a little ambiguous about this. He said "So I asked: What would I have said?" It is not clear whether he asked himself or ChatGPT.

It appears pretty clear that PG wrote what he actually thought in response to ChatGPT's render of his potential thoughts.

The thing is, I have to trust that PG wrote the words saying he wrote his own response. What stops ChatGPT from saying those same words?

I'm wondering what ways/technologies we'll create to tell if someone actually generates text themselves. Right now, I can feel a rising desire to avoid text-based communication with people, almost a fear of getting catfished by text.

That is an interesting concept in itself, if the idea is sound, why does the originator matter?

Usually if one wants to take advice, one should consider the reliability of the source of advice as a factor.

Most certainly we do, but I find the question interesting, a focus on advice verifiability helps both cases.

Yes, but is the idea sound? ChatGPT can produce fine-sounding nonsense at least as well as a human can. The difference is that I have a lot more experience calibrating my BS meter with humans than I do with ChatGPT.

Reply on news.ycombinator.com