Comment by TimTheTinker
8 years ago
> But we do know there must be algorithms capable of solving the Turing test, since there are only finite questions you could ask and only finite answers you could get back
Only if the algorithm itself encodes the entirety of those finite questions and answers. You might as well say that pi (or any irrational number) is intelligent, since every sentence ever spoken by any human is encoded therein somewhere.
The problem is that intelligently selecting the index of pi to use at a given time (or writing the algorithm that knows what to say in advance) is an act that itself requires a human mind's intelligence. You're only moving the problem one step back. You might as well reverse engineer every possible AES-128 output for every possible key and input (up to a certain length) and declare the cipher broken. It might be possible given infinite computing resources, but doing so wouldn't offer any insight.
Having insight is different than solving the problem. For example, developing AlphaGo didn't give us any insight into how to play Go (besides through watching it play). Similarly, it might be the case that developing a software which passes the Turing test gives us no insight into the nature of language. Nonetheless the problem would still be solved.
> You might as well reverse engineer every possible AES-128 output for every possible key and input (up to a certain length) and declare the cipher broken.
But we don't say that AES is impossible to crack. We only say that it's infeasible to crack. Just like with Go, where we had algorithms to solve it even 50 years ago, but it was just infeasible at that time. If what you are saying is that solving the Turing test is possible, but infeasible, then that is a more agreeable position to me. Regardless, I don't see how you can say that the Turing test is "spiritual in nature" while AES isn't.