http://www.the-brights.net/forums/forum/index.php?s=&showtopic=7643&view=findpost&p=124254
brightskeptic @ Nov 28 2007, 23:30
So, the question of whether an AI can be conscious is, can a computer (system of hardware + software) produce behavior that can reliably stimulate the consciousness judging modules found in normal, healthy humans. That is how I define consciousness.
This doesn't seem like a definition to me - it seems like you are defining consciousness as being the behaviour that stimulates my consciousness judging modules (which are probably fabrications).
Actually's Ron's statement, like the TT itself is not a definition but a statement of operational equivalence. It functions as a definition when all else fails.
But the TT-AI would certainly be considered conscious in my def. The "environment" of a bbs gives the astronomical+ scope of language as both options for interaction and "mental state". And the coherency and direction of discourse implies "intention" and thus a prediction of acquiring some additional mental state. In short to pass the Turing Test an AI would have to fall within the human norms in the dimensions implicit in my definition.
But the definition is important because the TT only gives a positive if the mental states and inferred intentions are also very similar to those human. In the process of constructing an AI, though the human mind is likely to be the first working model, we are (extremely) likely to find mappings of "mental state" and predictive responses far outside the range of human behavior. Likewise should we ever find life off earth, particularly if its time-response differs radically from ours, we need to be able to identify and classify "consciousness" which is radically different from our own.
-- TWZ
|