Ask HN: Do you think Gemini 2.0 Flash is using humans to help pick responses?

[veeter]

I can't get this idea out of my head.

Sometimes it's definitely near instantaneous in it's responses, so there is surely no human behind the curtains.

But other times it's literally taking 30-45 seconds to even begin responding at all. This happens at all times of the day so I doubt it's differences in load capacities.

I honestly think they have a team somewhere hand picking some of the best possible answers.

You know how ChatGPT asks you directly "which do you prefer?" ... could Gemini be doing that with our responses?

[rolph]

using humans to pretend the AI works better than actual, is how you generate serious concerns for investors, and shareholders.

[veeter]

It's deemed "experimental" so presumably they could do anything to augment it.

[SunlitCat]

(Not quite serious, tho)

Actually, if you have a whole bunch of click workers doing that work for you...hmm.....

Wouldn't be funny if that's really what AI is? A bunch of people coming up with answers to questions you give them, which would explain why it sometimes answers questions and sometimes it declines to answer questions or just says it can't? ;)

(Yeah yeah, not really serious here, just kidding :)