Provide a “captcha” prompt, and ask which comes first in the list… You know, to make sure the user is a human! And you get free sorting.
I mean, this strangely is a compelling idea…
I always try to prove myself human with wrong answers. Typically the first screen is an AI training screen, then the next screen is one the computer knows but wants to see if you know. By answering the first one wrong and the second correct, I’m hoping to poison the data.
Iirc they require 5 votes with the same response. Your effort won’t do anything, other than flag you as unreliable.
Afaik 4chan (of course) did that poisoning correctly, “classifying” the unknown words as “penis” (of course).
I’m happy enough being classed as unreliable, either way they don’t get good training data.
