[nickjj]

I'm certainly not holding my breath.

In a handful of prompts I got the paid version of ChatGPT to say it's possible for dogs to lay eggs under the right circumstances.

[SoftTalker]

Do you believe you could not find humans who would do this?

[raddan]

That's not really the point. If our definition of AGI does not include "being able to reliably do logic" then what are we even talking about? We don't really need computers with human abilities--we have plenty of humans. We need computers with _better_ abilities.

[NoMoreNicksLeft]

His objection might be that those humans aren't actually intelligent.

[fritzo]

I've long been terrified of the existence of adversarial prompts that can get me to say anything, that dogs can lay eggs, that there are five lights, that here's my bank info

[fragmede]

That's the terrifying thing about ASI. It could convince me, using everything the AI has on me, from all of my digital footprint, to do whatever it wants me to to, just by saying the right thing to me in just the right way, by copying the voice of everyone I've ever talked to, and by sending a humanoid robot in a skin suit that looks like them to my house.

I give it 10 years, maybe, for that to exist.