[extropy-chat] will the sun rise?

[Eliezer Yudkowsky]

Eugen Leitl wrote:
> On Thu, Dec 23, 2004 at 02:49:58PM -0500, Eliezer Yudkowsky wrote:
> 
>>So you believe that it would be physically impossible to develop stellar
>>disassembly technology within 10^51 Planck increments?  I suppose you feel
>>that, say, 10^52 Planck increments or 10^53 Planck increments would be more
>>realistic?
> 
> Control disappears well before Planck scale. To the best of our knowledge
> machines are restricted to condensed phase between 0 and a couple of kK. That limits
> the autoamplification rate quite noticeably.

Mm... if we assume that our model of physics is not going to change 
significantly in the 21st century, as it did in the 20th, then I agree that 
operating with nucleons in the kiloKelvin range would prevent you from 
completing operations at Planck speeds.  It would still run faster than 
neurons, though, so the subjective feeling that "20 years is not a long 
time, while 2000 years is a long time" is just as silly a guide to our 
expectations of how long it takes to develop solar disassembly tech.

Actually... I'm not sure I agree with this, if you're talking about 
cognition rather than motor actions (see below).  Maybe you could run a 
photonic computing operation closer to Planck speeds, though it had better 
be reversible.  I don't see how you could actually get within ten orders of 
magnitude of Planck speeds, but you could get closer.

I am just making the point that from the perspective of physics, a human is 
a glacier; a gigantic, motionless monument.

>>Never forget that you run on a 200Hz processor.  Your timescale is not the 
> 
> You don't "run" on a "processor", and it's most assuredly not "200 Hz". It
> doesn't matter how fast you can think, but how you quickly you can move
> stuff. Without breaking your machine which does the moving.

Are these not essentially the same problem?  Well, maybe if you presume a 
pre-existing object to operate on, like the Sun, then you could compute 
faster than you could change the pre-existing system.

>>timescale of physics.  Unless you have set forth the physics of stellar 
> 
> You have to autoamplify and get out the gravity well to next big pile of
> atoms (predispersed material in shallow gravity wells, preferrably, and these are many lightminutes
> away) and disperse that into a large circumsolar machine. This is readily
> feasible within 20 years. I can't put a lower limit to it, but it could be a
> couple of years, and maybe even months (I doubt latter, though).

I don't see why you would doubt the months, if you rushed as fast as 
possible.  Transit time from asteroid belt to solar system of days is 
feasible if you are willing to expend lots of fuel, same for putting gross 
materials where you want them and transforming them into machinery - 
presuming that is how you are disassembling the Sun.  It's a question of 
how much energy you're willing to waste.

> Assuming you could catalyze matter-energy conversion (we have no idea this is
> ok with physics), would asymmetric explosive stellar ablation (with most material
> lost into space) qualify as disassembly?

One suspects that the primary issue would not be speed as such, but 
choosing a balance of speed, power expenditure, and material lost, which 
minimizes entropy loss and hence maximizes the calculations performable 
with the winnings.  Does it take more energy to rush machinery into solar 
orbit, than the Sun wastes over that time?

-- 
Eliezer S. Yudkowsky                          http://singinst.org/
Research Fellow, Singularity Institute for Artificial Intelligence