"A trivial task for humans"
Not really, we spend years learning how to pick stuff up, just like how we spend years learning our first language or to just walk.
And the above is within a system where the generic feedback loop of failure and success is a solved problem unlike in the field of AI.
Picking up a cup of tea is surprisingly difficult when you look at all the actions required.
1. Move your arm to exactly the right place - one inch any way and you're either missing the handle or knocking over the cup.
2. Figure out how many fingers will be comfortable in the cup handle and insert them into the hole - often there's much less than a centimetre of leeway.
3. Exert the right amount of pressure between fingers in the handle and below the handle to keep the cup balanced and level while it makes the perilous journey between mat and mouth.
This is only possible due to how astoundingly complicated and sensitive our nervous system is as you have a good idea where any part of your body is even if you can't see it due to the stresses and strains on the rest of your body which your brain then interprets.
That's not to say those pulling all nighters via egregious use of coffee can't do all these actions on autopilot, but it did take literal years to get the ability to figure out all the steps for themselves.
Biting the hand that feeds IT
