That would be weird. Some of the tasks on there now are for labeling training data for AIs ...

Might be worth it if you have a super smart AI that is very slow to use its predictions to train a very fast highly specialised neural net.

Also, the whole thing makes me think of the original plot of The Matrix - the one where humans are used as co-processors rather than very inefficient organic batteries. The Wachowskis stated that at the time they thought people wouldn't understand the former, so they went with the latter. Today I can absolutely picture farms of people born in pods living in virtual reality where everyone has the equivalent of an AMT job and nobody finds it weird that you're paid to solve made-up problems because duh, everyone does that. Maybe at first 80% of people would work in some sort of management job so the machines can learn how to manage people and then gradually displace them.

No you misunderstand, this is horizontally scalable humanity. This is how you bootstrap training models.

Would their new slogan be Artificial Artificial Artificial intelligence?

That's almost certainly Amazon's endgame.

Don't know if this would really happen much, unless we are saturated with labelled datasets. They are used for labelling and not directly responsible for the actual end classficiation which is what AI is aiming to solve. In fact, AI needs labelling to work on these classficiations

classfic[i]ations

/Just a human doing a spellbot's job.

class[i]fic[i]ations

/Just another human doing a human's job doing a spellbot's job

Makes you wonder about the quality of the job done on MechanicalTurk.

Redoing supposedly done work is my worst nightmare.

Haha this is priceless. :-) Thanks