how do we evaluate intelligence, either real or artifical? are there sufficiently objective metrics for this task. attacking the definition of intelligence from a computational perspective, is easy to define on many objective dimensions including
speed, memory, precision/recall. unfortunately/fortunately the objective criteria do not necessarily give real intelligence an edge over artificial (i.e., machine) intelligence. of course we have more fuzzy criteria like algorithmic ability, creative
potential etc.
from a computational perspective its nice to believe that on purely objective criteria there is no distinction between real and machine intelligence, but from a human perspective its hard to accept since we can always come up with many subjective metrics on which machine intelligence comes up a cropper. if we have a battle between human and machine intelligence, both sides would never reach a consensus on who would win, because the machine intelligence does not accept the human criteria as valid metrics since it cannot objectively understand or interpret them, while the human intelligence will claim victory since it can always create more metrics on which it can claim superiority.
maybe never the twain shall meet ... and we can all have a good nights sleep without worrying about the toaster turning into a tactful adversary tomorrow morning.
Subscribe to:
Post Comments (Atom)
1 comment:
follow-up thoughts .. some of the seminal work in cs has been essentially in quantifying different aspects of human-intelligence -- computation, complexity, information theory etc .. it seems odd that no one has tried to quantify intelligence but have instead sought to expend a lot of resources in creating artificial intelligence .. it would seem in hindsight that the first half-intelligent step would have been to come up with some quantitative evaluation of intelligence .. maybe its this absence that is killing that community of researchers ..
Post a Comment