Government Regulation Of AI

The following is copied from Wiki Interview With Eliezer/Issues Of Concern:

Are you concerned with the possibility of strict regulation on AI? If so, why?

Perhaps I am overly pessimistic, but I see absolutely no hope that a government committee could legislate Friendly AI. Friendly AI is a basic research issue of equal depth with General Intelligence and seed AI, not as large, but just as deep. If a government committee cannot build an AI, they cannot figure out how to make it Friendly. As for the "relinquishment" of AI, Bill Joy's euphemism for bans, as far as I can see this will either transfer the development of AI to rogue states and organizations, or else delay the development of AI until one of the other Existential Risks wipes out humanity. I am concerned with the possibility of the regulation of AI because the expected outcome of such regulation is catastrophe.

Does the possibility of a call for relinquishment on AI development, as has been done with cloning, affect your present-day decisions and actions?

Because there are many people acting in the realms of politics, and even a substantial number of transhumanists doing so, my action in this realm tends to be stating the argument against government control of AI design features, and occasionally minor things like posting the first comment in the Reason debate between Stock and Fukuyama.

If you believe regulation of AI in the future represents catastrophe for us, then what do you propose is done to prevent said regulation?

I believe that if the creation of AI is important, then given my skills and the existing balance of investment, most of my efforts should go into accelerating AI, rather than political opposition to the regulation of AI. It seems unlikely to me that AI will be banned in the next five years, but it is possible - the question for me is not "likely" or "unlikely" but what I can do about it. I suppose that I would recommend that others carry on the fight against regulation of genomic technologies, and nanotechnologies if applicable. AI currently does not appear to be on the radar screen, but I suppose I would recommend fighting against current efforts to mandate features in all computer hardware and software. I think the initials of this particular evil plot were SSSCA (Security Systems Standards and Certification Act). In the end, the government is too large to push on and too blind to persuade, so while I do think that opposing malregulation is important, I would not count on it working.