In 50 years, the machine will be able to outperform the bottom 50% of the human population in any conceivable endeavor or task behind a keyboard/monitor. So CAPTCHAS will eventually be useless.
So it's an intelligence race. The more intelligent system (a computer, human, or hybrid) will be able to deceive the lesser intelligent system (computer, human or hybrid).
So the problem changes from: "Is the entity I am dealing with a computer or a human" to: "Does the entity I am dealing with have intentions to gain an unfair advantage over this transaction."
Humans have a part of our brain where we look for how others might be taking advantage of us, the algorithms and data structures there are remarkable. Computers will need to acquire those abilities if they are to manage the exchange of money, goods and services.
Captchas are really just an efficiency measure. You could simply have a human screen every single comment (or signup or whatever) and cull the spam. But human time is too limited, so we come up with lesser measures.
When machine intelligence advances that far, we'll be able to easily automate human-level examination of every single comment or submission or whatever. So I think the problem will get better, not worse, with that level of power.
So it's an intelligence race. The more intelligent system (a computer, human, or hybrid) will be able to deceive the lesser intelligent system (computer, human or hybrid).
So the problem changes from: "Is the entity I am dealing with a computer or a human" to: "Does the entity I am dealing with have intentions to gain an unfair advantage over this transaction."
Humans have a part of our brain where we look for how others might be taking advantage of us, the algorithms and data structures there are remarkable. Computers will need to acquire those abilities if they are to manage the exchange of money, goods and services.