Friday, December 5, 2008

The Prospects for the Computronium Universe


This is the first of several blog entries addressing the far future - the remote, steady-state future of the human race.

If you think the near future is subject to flights of fancy, that is nothing compared to the incredible leaps of fantasy swirling around the far future. Here is just one, Ray Kurzweil's favored vision of the post-Singularity Earth:

1. At some point, transistor size will no longer be able to shrink. This theoretical minimum size of "transistor" (or whatever form that atomic unit of physical computation takes) is called "computronium". Computronium is also the nomenclature used for macro-level objects that are composed of these atomic units (what we call "chips" now).

2. However, since we will presumably need ever increasing amounts of computational power, the only alternative to provide these computational resources is to start converting more and more physical matter into computronium.

3. First, the Earth would be turned into computronium. Next, the solar system. Once the entire solar system has begun to earn its keep by becoming "thinking matter", our future selves will proceed to convert eventually the ENTIRE UNIVERSE into this substance. This is called converting "dumb matter" into "smart matter".

Kurzweil is absolutely brilliant in some ways, and the fact that this is the predominant vision for the far future proves this. However, this is brilliant visionary marketing only, and there are problems aplenty with this bold idea from a scientific reasonability perspective:

1. For starters, this is easily the most ridiculous thing I've ever heard in my life, and for Kurzweil to serious propound this as an obvious future tests my respect for him deeply.
Why is this apparently bold concept ridiculous?
Let's step back for a moment. As an inventor, Kurzweil is a genius. As a futurist, however, he is essentially an aggregationist, borrowing ideas from science fiction writers (this is where the Singularity comes from), the speculations of various thinkers, and stitching them into firm predictions. The computronium universe idea comes from Jim Gardner and the armchair speculations of physicists regarding very-large scale computational structures, such as the Matrioska Brain, a mega-computer that entirely encircles the Sun (not unlike a Dyson sphere), grabbing all the sun's energy for its computational power requirements. The computronium universe is an extension of these armchair speculations, which were never intended as actual engineering projects by these physicists.

2. As I've noted at some length on various boards, this is the cart before the horse. We currently have absolutely no idea what problem would require such a monster computational engine. Estimates vary, but the computational density of computronium is vast - something like 40,000,000 Pentium IIs on a chip the same size as the ones we're familiar with today. This would allow an individual to have almost infinite computing power in a footprint the size of an iPod . Therefore, scaling this up to the size of asteroids, planets, and the entire universe is an amount of computational power that we have no idea how we might use. Until we do, it's best not to predict the solution when we have no idea what the problem is, and extrapolating certain trends (such as increasing computational use) while ignoring others (such as innovative architectures that keep the footprint of computers constant or shrinking) is not empirically responsible. It's cherry picking the data to support a notion that was predetermined, apparently for its dramatic or elational appeal.

When I have asked the question (and I've asked it many times), "what problem are you trying to solve with a computronium universe?", no one really has a clue, Kurzweil included. However, the reason I hear most often is "in order to simulate the universe." So, the upshot is that we will destroy the universe in order to simulate it. To be polite, that seems unnecessary.

3. There are other points working against this idea. For one thing, as any planetary scientist will tell you, matter is not "dumb" - it is full of stories, full of history, full of insights as to the evolution of the universe, and this holds true if you're talking about a small asteroid or comet, on up to galactic mega-structures. To fully understand and appreciate these stories requires more than a snapshot before you destroy it - er, I mean make it smart by converting it to computronium. There is the element of dynamics, behavior over time, to understand the full complexity and information that that object or objects can impart to our still nascent appreciation of the mysteries of the universe.

4. In any case, an advanced civilization that went around sterilizing everything it encountered into a "thinking" substrate would make the Borg look like Mary Poppins. I would take my chances with the goony aliens from "Independence Day" any day, over the nightmare civilization that Kurzweil happily predicts.

When you go this route, you run into some problems. For one, we are probably not the very first technological civilization to ever exist in the entire universe. I believe technological civilizations are quite rare for reasons i will explain, but to presume that we are the only one is as irrational as assuming that they populate every star system.
Since we are a young technological civilization, any others out there are probably much older. That means they would have encountered their Singularity long ago, and their computronium-building programme must have been well under way or even completed long ago. But, where's all the computronium then? Rather than question the computronium idea, that is left intact, and resort is made to speculations such as "we must be a simulation inside a computronium universe".

Another point that must be invoked to support the computronium universe is that to turn the universe, some 15-billion light years across (that's just the part we can see with our telescopes), implies that we can reach all these remote locales in a reasonable time frame. Kurzweil uses the Singularity as a magic wand to wave away known physical constraints on just how fast physical objects composed of matter can travel. Now, relativity predicts and observations have proven the reality of mass dilation - that is, physical objects increase in mass as they approach the speed of light. Even a speck of matter as tiny as a proton, if it were to attain the speed of light, would increase in mass to a point where it would exceed the size of the known universe. This is one of the strangest phenomena I know of, but it is not speculation - increases in mass for particles in particle accelerators attaining a significant fraction of the speed of light increase in mass to the exact degree predicted by relativity. While I am not stating that it is impossible for this to be overcome in some way at some point, to simply wave it away with the magic wand of the Singularity is highly irresponsible. It is far more likely that this basic physical law will hold, into a future far too distant for anyone now alive to see. I don't claim many things are "impossible", but controlled, safe, faster-than-light travel for macro-level objects like humans and spaceships (or even protons) are one of a handful of things I would comfortably put into the "impossible" bucket, and empirical evidence and theoretical foundation back me up 100% on this.

I have a different view of the far-future. It may not seem glamorous, but I'll lay money down on it as a far more likely framework for the real future than the ideas discussed above. I will present these ideas in a future blog entry.

Further Reading:
Kurzweil's case: Forward to the Intelligent Universe

Wiki Definition of Computronium

3 comments:

teknoarcanist said...

I agree that the idea of converting the entire universe is a bit too large and abstract to consider seriously just yet.
I would take issue, however, with the idea that we would have to 'destroy' matter in order to convert it into an aspect of a thinking system. It would seem to me that if we have the ability to use matter as the rough equivalent of a binary digit by that point (or a transistor, whichever metaphor works) then we would have sufficient mastery of the physical universe as to do so in some way that wouldn't need to 'destroy' it, or even convert it to some super mystery substance. I sort of see it as we would simply be taking control of the atoms inside, say, a rock, and using them for computations since they're not doing anything besides being rock atoms. I don't imagine we'd have to turn them into some kind of amorphous cloud of goop to do this.
God, arguments at this level are so pointlessly abstract and arbitrary XD

teknoarcanist said...

woops, forgot to add:

That having been said, if it weren't necessary to 'destroy' matter to use it as a computational device, who's to say it's not being done? Not that we're a simulation inside a computronium computer ("DAAH I LIKE TEH MATRIKS"), but that the matter which comprises us isn't being used AS one?
But then that goes back to 'what question are you trying to answer', which I agree is one nobody seems to know the answer to with more depth than "LOTS OF COMPLEX THINGS THAT DON'T CONCERN YOU SHUT UP"

Justin said...

One reason to want such computational power would simply be to maximize the number of people (e.g. http://www.nickbostrom.com/astronomical/waste.html ). I don't personally see a problem with utilizing insentient matter this way.