> Jeff Davis <jdavis@socketscience.com> wrote:
> removed
...entertaining & realistic story about one probable path to AI...
> So the AI either stays locked up until it's really and truly socialized
> (boring but safe), or we hope that in its first self-liberated round of
> self-enhancement it jumps immediately to forgiveness and tolerance (klaatu
> barada nikto).
> I seem to have painted myself into a corner, and I don't like stories with
> unhappy endings. The government at its best would be a poor master for a
> superior intelligence, and the spook/militarist/domination-and-control
> culture is hardly the government at its best.
> So, my futurist friends, how do we extricate ourselves from this rather
> tight spot?
I think it has been mentioned already. Open-source.
If AI@home gets off the ground it is likely to bypass any efforts by the military/government. Question: Is the processing power being used by SETI@home now greater than that being used by the NSA? Perhaps not, but I suspect it isn't far off. It becomes a simple question of enrollment. The government is unlikely to enroll the population as a whole into devoting its computers & electricity to AI self-evolution (for the purposes of the government). I doubt that the government could construct a computer bigger than one that a few million private individuals could collectively piece together. The cry of the Z generation may be "I want my pet AI and I want it now!". Think of all of those Nintendo/Sony game-stations connected by fiber and co-evolving an AI.
However, there is a thread of darkness here as well. If AI@home does succeed it will be very difficult to control. It only takes one individual to take the safety-interlocks off and the AI becomes self modifying. As I pointed out in a much earlier message, the moral thing for the AI to do is to eliminate its creators since humans cannot be engineered/guaranteed to be "trustable". Of course self-modifying AI's could fall into that category as well. The situation is resolved if there is only one AI. On a good day I'm pretty sure I can trust myself.
Robert