RE: Jupiter brain substrate (WAS: Goals)

Billy Brown (
Thu, 20 May 1999 14:31:31 -0500

Eliezer S. Yudkowsky wrote:
> Well, I suppose neutronium is *okay*, if you're not worried about
> *efficiency* or anything...but Real Powers Use Monopolium. Not
> neutronium. Not Higgsinium. Raw, incredibly dense monopoles.

Alas, it appears that monopoles don't exist. And now, stay tuned for our next episode of 'As the Standard Model Turns'....

But actually, I was thinking of a high-density soup of many kinds of subatomic particles (see, I'm vague enough to survive the next annual revision of the particle zoo). You tailor the mass of the underlying gravity well to give you the correct pressure/temperatur regime for your preferred mix of particles, then build the computer on its surface. I'd expect it to end up being some weird mix of exotic particles, half of them things that we don't know about yet.

> Thus an actual Jupiter Brain, formed of Jupiter, would probably be just
> around a meter wide (?). You can't have a Brain that's actually the
> size of Jupiter because it would collapse and form a black hole, not to
> mention that there isn't that much mass in the whole solar system or
> possibly the galaxy..

Hmm. Maybe 2-3 solar masses per computer, with a probable diameter of 10-30 km each. How about we put a few hundred of them in close orbit around each other, in a volume of space a few tens of thousands of miles across? (Computing the trajectories will obviously not be a problem.) Enclose the whole structure with a shell a few light-seconds across (and a few hundred km thick, with lots of firepower), with a few solar masses worth of robotics on the outside, and you've got a pretty nice SI body. The total budget is only around a thousand solar masses, so we can support a substantial population just by dismantling the Milky Way.

> Questions for Thought and Study:
> 1. What is the maximum size/mass of a JB constructed of monopolium?

I don't think we have the physics to compute this - we'd have to know exactly what kind of environment we needed to maintain to make stable structures possible.

> 2. How much computing power would it have?

Well, let's see. Based on Drexler's work, we might estimate nanocomputers at something in the general neighborhood of 10^14 MIPS per cubic cm (for electronic computers with no special benefits from expiating quantum effects). For a Jupiter brain 100,000 km in diameter, with 10% of its mass actually used for computers, that gives us about 10^43 MIPS. It takes about a second for an electrical signal to cross that distance, so the time required for information to propagate through the entire system should be <10 seconds.

If we increase the component density by a factor of 10^16 (taking a middle-of-the-road estimate), switching to degenerate matter boosts our speed to 10^30 MIPS per cubic cm. For a computer in the form of a thin shell covering the surface of a neutron star, 10 km in diameter and 10 m thick, that would give us 10^45 MIPS. The communications lag drops to <1 ms. More optimistic assumptions would give much greater computing power (10^49 MIPS for a 1 km thick shell at your 10^18 density figure, for instance).

> 3. How much time would a unified thought take to cross the JB?

Maybe a few seconds for the nanotech brain, or a few milliseconds for the degenerate matter version. Of course, the appropriate thing to compare that to is the time required for all of human civilization to hear about and understand a new idea. A cubic centimeter of the picotech machine could support the equivalent of 10^22 human minds, with a communication lag of a few microseconds. A human-equivalent mind (in a VR) could easily experience a time rate billions of times faster than ours.

> 4. How does a JB's reaction time and flops/thought compare
> with a human?

Well, Hans Moravec pegs the human brain at 10^8 MIPS. That means the nanotech brain has 10^35 times as much computing power, and the picotech system has 10^37.

> 5. How does strange matter compare to Monopolium?

If only we knew.

> 6. Adjust your results for relativistic slowdown due to the JB's mass.

Not really a significant effect in either case.

> 7. Calculate whether monopolium has a Chandrasekhar limit. Does this
> change the maximum?

Everything does.

> 8. How would you detect a JB using SETI?

That's easy. If you have a SETI program, there aren't any JBs out there.

> 9. How much energy does a JB dissipate while performing computations?

Urk. Umm - we're going to be measuring that in Sols (1 Sol = the energy output of the sun). I don't have a handy estimate for this - does anyone else? We'd better put a black hole in the middle of that JB cluster to power the thing.

> 10. What is the maximum non-hole-forming density of JBs as a function
> of population?

Not easily estimated, since you have to trade off between lags due to physical distance and lags due to relativistic slowdown - those orbits are going to be rather fast.

> 11. At the population equaling a galactic mass, what is the time for
> thoughts to cross?

Using a JB clustering arrangement like the one above, you could easily fit a few dozen individuals into a space the size of the solar system. Space these groups closely and you can probably condense the whole galaxy into a few hundred million individuals living in a space a few light years across. That would give you communication lags of hours to years (depending on who you want to talk to), and atrocious crowding.

> 12. How does the ratio compare to human thoughts vs. human speech?

Obviously, the JBs are far more isolated.

Billy Brown, MCSE+I