On Thu, Aug 23, 2012 at 9:58 AM, [email protected] <[email protected]> wrote: > > Some might say, better a known enemy. Anyway, why all this stress on > self-modifying AI? Wouldn't it be easier & safer to design an AI that > doesn't want to modify itself than to design one that's supposed to stay > friendly despite ongoing self-modification?
The safest AI would be one that doesn't want anything. It would have no goals and no motivations, no reward button and no utility to optimize. It would be a vastly intelligent tool, a collection of all the world's knowledge and the computing power to do whatever you want with it. Rather than think for itself, it would be an extension of our own brains; a place to store your memories, communicate with anyone on the planet, and do the work that you would if you knew more and thought faster. It would be collectively owned, controlled by no single person but by everyone that uses it. It would be the AI that we are actually building; the one in front of you that has already surpassed human level intelligence in all but a few domains as it doubles in size every 1.5 years. -- Matt Mahoney, [email protected] ------------------------------------------- AGI Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-c97d2393 Modify Your Subscription: https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-2484a968 Powered by Listbox: http://www.listbox.com
