RE: SIAI's flawed friendliness analysis

Date: Mon May 26 2003 - 17:19:43 MDT

In MY opinion, a massive, well-funded, mistake-free, government-supported, single, unopposed AI project would be very nice. Unfortunately, it's not going to happen that way any more than I can grow pizza on my head. Anything that gets tax dollars - never mind full government funding - gets bogged down in the swamp of lobbyists, regulations, and political debates. Just look at the tax code - it now would fill a full set of encyclopedias. Second, 90% of the public isn't going to have the slightest idea - or be interested even if they did have an idea - of what AI is. The public is just too busy going about their daily lives. Third, advanced AIs can't be regulated. Period. We have a hard time regulating Windows XP - let alone an intelligent being. AIs are simply too sophisticated, and they are too GOOD at things like obscuration and programming code, for a team of merely human regulators to do much good. The terrorists had no trouble pulling off 9/11 - imagine what a being with a hyper-fast computational brain
could do! May I quote Eliezer - "Humans stink at network security; it's not our native environment." That's what I call the Home Team Advantage - people in their native environments can't be subdued, or even regulated and inspected, easily. Our native hunter - gatherer environment is eons away from modern society - an AI's native environment IS modern society. Since AI can potentially do good for everyone, everyone will want an AI to do their bidding. Regulating AI is like trying to regulate nuclear weapons, worse, intelligent nuclear weapons, who are ten times more skilled about getting along in our modern day environment and manipulating it than any human, and can disguise themselves as ordinary humans with good intentions.

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:42 MDT