| Hard problem, simulation argument |
[22 May 2013|10:49pm] |
This is kind of a dumb idea. It may actually qualify as my first "stoner philosophy" post. Anyway, I got started thinking about it after checking out the simulation argument last year.
For the sake of argument, let's say you have unfettered ability to simulate and observe possible worlds for the purposes of evaluating different forms of artificial intelligence. You want to know if an AI which passes the Turing test also possesses consciousness (I think I am correct in assuming that an AI could pass the test without a shred of consciousness in the "hard" sense).
So you run a kind of Borgesian social simulation; your AIs are not interacting with humans, but with each other, in large numbers, preferably over a very long period of local time -- centuries, let's say. Their little AI civilizations rise and fall, produce their own mathematics, art, and literature, and so on, with no access to information about the "outside".
My question is: would it mean anything, if after running this simulation a million times (let's say), none of the AIs -- in their attempts at philosophy -- reproduced the problem of consciousness? Or anything that could reasonably reframed as the same problem? I.e., could a non-sentient intelligence problematize a feature of mind which it does not possess? (One might also wonder whether a non-sentient intelligence would philosophize at all; I assume so -- and it would look just like analytic philosophy.)
My own intuition is "probably not", but I'm also unhappy with my hypothetical. The main problem I see is that it's very plausible that a sentient intelligence could simply ignore the most obvious feature of its own mind (Dennett and plenty of others can, apparently).
|
|
epilogues_