Re:Peter Norvig critically reviews AlphaCode's code quality
by pishpash on Friday December 16, 2022 @07:58PM
(#34022932)
Are you sure there is an actual "thought-process [sic] of humans"? Witnesses have made up explanations for false memories, the same as in a dream.
Re:Peter Norvig critically reviews AlphaCode's code quality
by galaxyLogic on Friday December 16, 2022 @08:08PM
(#34023026)
I am not sure. But humans can often explain the thought-process that led them to a certain decision. I think that counts as evidence of consciousness., in humans.
Re:Peter Norvig critically reviews AlphaCode's code quality
by skirmish on Friday December 16, 2022 @08:25PM
(#34023189)
Humans can explain rational thought, sure, but not subconscious or acting on impulse. When you say something you shouldn't have said, how do you explain the thought process that led you to saying it? Were you an unconscious actor when you unwisely started an argument with your manager?
Re:Peter Norvig critically reviews AlphaCode's code quality
by galaxyLogic on Friday December 16, 2022 @09:51PM
(#34023883)
Yes you can be conscious about having been unconscious earlier. We do have reflexes. But we also have consciousness about our own thinking.
Now mimicking the speech of people may give the impression that the AI has some thought-processes behind its speech because, it is text that any human might write. But there's no logic because there's no symbolic reasoning. There's just mimicry and trial and error. "Training the model" just means millions of trial-and-error practice runs.
What the (neural network) AI can NOT mimic is the actual (symbolic) thought-process of humans, because that is not visible anywhere. It can only mimic the input -> response -behavior of humans. Therefore no, the AI is not conscious. It is the proverbial zombie.