Hacker News Reader

Comment by HPsquared | Hacker News Reader

I've seen blatantly wrong stuff in that overview too many times, I just ignore it now.

To be fair, the actual results are often even worse. I'm pretty sure we're close to the point where our favorite AI prompt replaces classic googling. While it will get a lot of the answer wrong, it will lead to the right result faster than plain searches. If nothing else, because refining our search at the AI prompt will be way easier than in classic google. Google knows and needs to stay on top of this paradigm change, but I guess doesn't know how to monetize AI search yet so it doesn't want to force the change (yet).

Sometimes search results don't contain the info you need, sometimes they are SEO-spam, or a major topic adjacent to what you need to know about floods the results.

But they're not often confidently wrong like AI summaries are.

Ai used in this way is going to replace gui as we know it. Why click when you can just tell ai what you want to to do.

Because I usually don’t want to talk to computers in front of other people? It isn’t that it feels silly, but that it’s incredibly distracting for everyone to hear every interaction you have with a computer. This is true even at home.

Maybe we can type the commands, but that is also quite slow compared with tapping/clicking/scrolling etc.

I recently typo'd something and the AI box just fabricated a semi-plausible story about how the ancient X did Y.

I am not sure how everyone is ignoring that or just getting along with it. I've had it return that the earth population is 10bn which is blatantly wrong and infuriating. It hallucinates harder than regular LLMs which are already bad for getting reliable answers.

Reply on news.ycombinator.com