RE: Proving What We Already "Know"
July 4, 2022 at 5:45 pm
(This post was last modified: July 4, 2022 at 5:48 pm by bennyboy.)
(July 3, 2022 at 6:51 am)The Grand Nudger Wrote: Is that what people use their human rights for now?
No, I don't think so. Maximizing influence requires a level of effort and dedication that only a few of the most ambitious can consistently muster.
A biological system has a folded-in goal: evolve has led to the goal of continuing to evolve. So long as we seem likely to be able to survive and fuck, and have children with at least that level of motivation and capabilitgy, the rest (degrees of altruism vs selfishness, level of intellect, motivation to reach higher levels of success etc.) seems to be fairly random.
An AI system works in the context of goals-- with failure or success to meet the goals in each attempt providing a mechanism for feedback without which the system cannot evolve. But those goals can be set arbitrarily by people.
Right now, an AI's goal system isn't turned in on itself. If it's set to learn a game, it won't (currently) consider that the best way to win at chess is to kill your opponent. But when it IS programmed to consider threats to itself (say by a military protocol), or to replicate (say by an attempt to terraform an Earth that is no longer what we want it to be and the need to produce mass-scale machines) then things could get interesting in unexpected ways very quickly.
