RE: Migrating SI (was Re: Yudkowsky's AI (again))

Billy Brown (
Wed, 31 Mar 1999 09:14:47 -0600

Dan Clemmensen wrote:

> You argue the the SI will not concern itself with slow atomic-scale
> activity because it doesn't want to wait. You postulate migration to
> a "swinging hot spot" instead.

Sort of. I think the kind of SI we were talking about is going to go so far, so fast, that it isn't going to bother with anything as mundane as large-scale nanotech. The actually options I listed are just suggestive - I expect a real runaway SI is going to come up with something much better.

> I don't think this works. If the SI is concerned with time, it surely
> want to take the time to migrate. It if wants a neutron star, it will
> convert the sun.

The sun isn't massive enough to form a neutron star. Their isn't enough mass in the entire solar system to make one. In more general terms, if high energy levels (in a cosmological sense) are required to build an SI's dream home, there probably isn't enough mass/energy in the solar system to support the project.

> If migration is the only option, the SI will take all the mass/energy it
> with it, probably making an optimum tradeoff between mass/energy expended
> for propulsion versus mass/energy that can be devoted to computation
> enroute.

If we assume that the SI has no morals, that there is no possible way of building what the SI wants without creating extreme conditions (in a cosmological sense), that there is no possible way of migrating to another location that would be faster than sublight spaceflight, that taking large amounts of mass along would actually make the trip more pleasant, and that the SI intends to actually experience the trip (rather than just switching to a waiting state for the duration), then yes, that might happen.

I think all of those assumptions are questionable, some of them highly so.

> FTL travel and customized universes require new physics.

Given that we currently know that our best theories of physics are incomplete, I think it is unrealistic in the extreme to expect that an SI will not discover new physical effects that it can exploit. There is no way to know what those discoveries will be, of course, but we must expect that they will happen.

I would not predict that an SI would discover any one particular 'impossible' technology. However, I feel safe in predicting that it will find ways of doing many things we would consider impossible.

Billy Brown, MCSE+I