I came across this issue on different forums where Google uses our answer when resolving reCAPTCHA to drive its artificial intelligence. For example, some say that spotting stop signs and cars in reCAPTCHA drives Google's standalone AI. I am not sure of the authenticity of these claims, but I doubt that it is even theoretically possible to train an AI this way.
ReCAPTCHA asks the user to locate or identify a certain thing and, by answering it incorrectly, asks you to try again. This means that Google already has the knowledge of the correct answers to the challenge. He already knows what pictures are stop signs and cars. Is not it enough to train an AI? Why is my contribution to the puzzle necessary to train the AI?