Hacker News Reader

Comment by al_borland | Hacker News Reader

I’ve been conflicted on this one, as I will ask it stuff I know a bit about and am looking at getting into the weeds on it, and it gives me bad info. Then I have to correct it to redirect it, and it’s eventually just a parrot. So can I really trust it to tell me what I don’t know at all?

It also feels like another form of brain rot… it’s like arguing online, but worse, since I know for a fact it’s just a bot on the other end.

View on HN

Can relate to this. I really wonder if the people using AI to learn new skills are checking the output against non-AI sources?

I’ve tried making training content for topics I know well, and LLMs will leave out very important information or write long pages about irrelevant content. The dangerous part is that it all looks and reads plausibly.

Absolutely true. On domains I am an SME it fails miserably and tells me popular half truths. But after steering towards correction it can be accurate.

I tell it not to be confident sounding if the conclusion is not strong.

Sadly commercial models have hidden "helpful guardrails" that are not helpful and want to "balance" answers with garbage.

Its an art to steer but it can be done.

I tell it to examine both sides or find flaws with a position.

I try not to lead it to my preferred conclusion.

Bad info can be mitigated by asking for human sources and citations.

Make the bot work for you.

Reply on news.ycombinator.com