That's not exactly what I said though. I didn't say "run yourself." What I said was, "Run this code with this input, and then do something different." A human could certainly do that. A strong AI should certainly be able to.
Hence it can't be defined in terms of code and input.
Not a problem for humans. My code and input are massively inaccessible. Modeling them, especially modelling all the input I've received--could involve modeling a big chunk of the universe. Whether or not they're even deterministic is an open philosophical problem. I'm pretty safe from being asked to "run" myself.
Not so for a simple Strong-AI chatbot, though. If it's just a program a computer can run, if you can enumerate all the input . . . then by definition, it can't handle a simple request that any human would be able to handle. Not strong.
Hence it can't be defined in terms of code and input.
Not a problem for humans. My code and input are massively inaccessible. Modeling them, especially modelling all the input I've received--could involve modeling a big chunk of the universe. Whether or not they're even deterministic is an open philosophical problem. I'm pretty safe from being asked to "run" myself.
Not so for a simple Strong-AI chatbot, though. If it's just a program a computer can run, if you can enumerate all the input . . . then by definition, it can't handle a simple request that any human would be able to handle. Not strong.