Are there any questions or tasks that are easy for a human but impossible for an AI?

that's the premise of the Turing Test :)

ton's of discussion/research/philosophy around that exact question: if we knew something that was easy for humans but impossible for an AI, then a Turing Test would be trivial.

Be specific.    
Building productscore.org, @germanify // leonasskau.co.uk // Hosting /geopolitics, /strategy, /leo
d33m:mun

that's the premise of the Turing Test :)

ton's of discussion/research/philosophy around that exact question: if we knew something that was easy for humans but impossible for an AI, then a Turing Test would be trivial. https://en.wikipedia.org/wiki/Turing_test

A doodler and computerer. I like permissive/permissionless open source, smart contracts, p2p systems, room-scale VR, and NixOS.

shazow.net

Currently: WhatsABI

Of course, but I’m not asking about an abstract system, I’m asking about systems that could exist in the 2020s

I’ve been thinking about things like “write tomorrow’s date on a piece of paper”, because that date might be outside its training data and be hard to extrapolate to

I think if an AI is not limited to a single inference pass, then it's hard to come up with something that can't be gamed, even writing tomorrow's date on a piece of paper.

IMO if possible it's better to ask what kinds of tasks/scenarios we really really care to differentiate AI vs human, vs what kinds of things don't matter?

Closest thing we may be able to do is to limit the number of AIs, but not whether something is an AI or not.