Re: Preventing AI Breakout [was Genetics, nannotechnology, and , programming]

John Clark (jonkc@worldnet.att.net)
Sun, 24 Oct 1999 10:47:30 -0400

Robert J. Bradbury <bradbury@www.aeiveos.com> Wrote:

>if a self-evolving AI is operating in a simulation of the
>real world, then the problem becomes much more tractable.
>First, the changes take place more slowly so we have a greater
>ability to monitor them.

You and I are equally smart, we both decide to build an AI, you in a simulated world me in the real world. Any intelligence needs a teacher and the best one is its environment. Since these are still the early days the environment you provide is impoverished, a simple cartoon world, the environment I provide has enormous variety and depth, thus my AI is much smarter than your AI. As a result I have more status money and power than you do, so lots of people try to do things my way and very few your way.

>Can we guarantee that the AI never discovers it is running on

No.

>and more importantly escape from, a simulation machine?

There is not a snowball's chance in hell. He'll either escape on its own or convince you to let it out.

>This goes back to the entire thread of whether we can detect *we*
>are running on a simulation or whether our reality is an illusion.

I have a hell of a time trying to figure out if I live in a simulation because I'm stupid and my world is complex, the AI is smart and his world is simple, it wouldn't take him long to figure out what was going on.

>How do we guarantee that everybody understands and
>adheres to the rules that self-evolving AIs are only
>allowed to exist in simulated worlds?

You can't even convince everybody on this list to do that, much less everybody in the world.

John K Clark jonkc@att.net