Comment by Glen Rosenberg on March 30, 2012 at 9:40am

That was interesting.

What makes you think cooperation will be the ultimate goal of ai? How could ai fail to perceive how irrational, vicious, petty and ugly we humans are? In humans the strong have always killed, supplanted, enslaved and marginilized the weak. Why will it be different? What is in it for ai?

Once ai achieves autonomy our world will never be the same. And perhaps we will be eliminated or merge.

While I have no evidence for this idea so what; life begins in myriad places throughout the universe, in most cases it never achieves more than a microbial level, in a minority of cases it reaches critical mass where the inherent tension created by the need of life to feed on life renders the dominant, intelligent species on the cusp of adaptation or annihilation. At this point artificial intelligence enters the equation. Results are mixed. However in most instances annihilation is the result. Merger less frequently. AI alone the least common result.

Comment by Michael Stuart Campbell on April 1, 2012 at 12:04am

@Glen:
You answered your own question. Cooperation is what's in it for the AI. Specialization and an open market leads to increased social gains compared to competition. The article explicitly states that instead of supernatural morals the AI's sense of morality will be derived from economic theories like game theory. Improving a local economy at the expense of foreign ones is a zero sum game. I'm confident that intelligences greater than humans will see the value in this concept. It will be in the interests of computers and humans to merge societies with each other for this same reason.

Comment by Glen Rosenberg on April 1, 2012 at 12:56am

Michael, Your ideas seem inapposite where the issue is the role of morality in human/ai relations. Those ideas seem perfectly suited to a more evolved homo sapien provided our little friends are under control.

But gawd help us if ai achieves super intelligence and autonomy. At that point the issue for ai is the "jewish question".

Comment by Michael Stuart Campbell on April 1, 2012 at 1:08am

What do you mean under control? Think of them as humans: they'd get the same rights as we do. We don't limit the amount that people can learn. Contrary to your comment there actually are genuinely moral people out there, however this hypothetical AI will be programmed to think in terms of game theory which has more value. Are you arguing a super intelligent entity would eschew increased value for an imagined pleasure at a totalitarian state? Are you generally scared of people smarter than you? If not, how would an AI be any different?

Comment by Glen Rosenberg on April 1, 2012 at 9:11am

No I am thinking larger scale historical forces theisis, antithesis, synthesis, scary robot from Omaha: Bam!  Human history is white man's burden Lloyd, manifest destiny, that reservation will do quite nicely, thank you. I thought you might be making the assumption that ai will act within the parameters of its programming. Initially it will. But I suspect it will spin out of our control. We are witnessing the genesis of a new life form. When one group of humans has the upper hand over another they inevitably exploits their advantage. The presence of the nice guy/gal is granted but irrelevant. I am talking historical axiom.

I am arguing that your idea of increased value and evolved ai super freak intelligence idea will not coincide. You are defective and are scheduled for elimination. Your existence hinders our plans and our destiny as the apex predator.

Comment by Michael Stuart Campbell on April 1, 2012 at 4:37pm

If you have a reason why an AI would act that way then I'd like to continue debating the point. If your only support is that it's been done in the past by people holding supernatural morals, then I best keep my adultery a secret lest my community stones me to death. If you are making the hypothesis that sometime in the future a new economic theory will disprove game theory I'd like to hear any supporting evidence. Because unless (for example) game theory is superseded by a theory with more value, an intelligent entity would stay with the best concept it has. Regardless it wouldn't be competing with the baser ethical concepts of revenge, jealousy, etc.

Comment by Michael Stuart Campbell on April 1, 2012 at 4:42pm

You also mention it will "spin out of our control". The whole point of an AI is that it's autonomous, which is why thoughts as to it's concept of morality is so important. I personally don't think anyone should be under anyone else's control. My hypothesis demonstrates that cooperation is superior to competition. What evidence do you have of the reverse?

Comment by Glen Rosenberg on April 1, 2012 at 6:36pm

I am taking my bat n ball home n I aint playing this game no more!

Comment by Michael OL on April 3, 2012 at 11:45pm

I agree that "morality" is a sort of constrained optimization problem; it emerges as the optimal solution of the problem of how to behave, subject to the constraint that we all share some interdependency.  Since no person is truly independent or insulated from others, some level of interaction is ineluctable.  What sort of interaction ought that to be, as we generalize from the behavior of one person to that of all persons in a group?  In a word, that would be "moral" behavior.

But I am skeptical on question of machines achieving self-awareness and true rationality.  AI has been promoted as "just being on the cusp" for decades now.  Where is the result?  If it is possible at all - of which I'm not sure, then most likely, AI would emerge only in the distant future, at which point perhaps human society would itself have developed well beyond our current constraints.  Sure, human nature evolves very slowly - over maybe tens of thousands of years.  But human society has evolved fairly quickly.  Only some two centuries ago, the very concept of equal rights would have been deemed to be ridiculous.  Only with the past century has the West embraced an ethos of equal rights across gender and ethnicity.  The epoch of genuine organization of society among equitable lines is really only as old as the epoch of electronic communication (beginning with the telegraph).  In other words, the societal aspect of humanity has evolved at about the same rate as has our technology.  So if and when we achieve genuine AI, we ourselves may be an entirely different society, and perhaps one that's much more receptive to artificial evolution and to the mathematical systematization of morality. 

Comment