Hacker News Reader

What are your actual prompts? And what model are you using?

This was Google Search's AI overview. But AI isn't much better.

So all this and not sharing of actual prompts for someone else to verify it on their end, rather than just hearsay. At this point I don't believe you.

What prompts? I merely did a regular Google Search, with typical keywords like I've done since 2004, and the automated AI overview showed a clearly wrong result

https://i.k8r.eu/1Pc-Tw.png

And why is that so hard for you to believe?

I'm especially surprised by this fixation on the prompts. You can't have it both, either AI is great because you can just ask for something, then the exact prompts shouldn't matter, or the precise wording matters, then I'd just want a regular, deterministic query language.

Wasting 5 days of back and forth just to get to this result is why I didn't believe you. But sure sometimes it's wrong based on what it picks up from random documentation, same as a human would be if they read something online that turned out to be false when they tried it out. As long as it's correct most of the time it's still useful.

Reply on news.ycombinator.com