Slashdot Mirror
Calculating A Theoretical Boundary To Computation
TMB writes "Lawrence Krauss and Glenn Starkman, astrophysicists at Case Western Reserve University (and in LK's case, author of a number of books including Physics of Star Trek), just submitted this nice little paper to Phys. Rev. Letters. It claims that in an accelerating universe, the existence of a future event horizon puts a fundamental physical limit on the total amount of calculation that can be done, even in an infinite time. This limit is much smaller than the traditional Hawking-Beckenstein entropy. Among other things, this implies that and Moore's Law must have a finite lifetime, here calculated to be 600 years, and that consciousness must be finite."
"consciousness must be finite"
;)
Except, of course, for those using certain popular mind-expanding substances
Seriously though - it seems we are finding a new limit every day. Wasn't it last week that they theorized limitations on data storage, as well as data transmission speed?
- To err is human; but to really screw up, you need a computer
Moore's law was never intended to be a scientific theory. It was just a useful observation. It has never had anything other than economic incentive to keep it going. Using it to discuss the calculational ability of the universe is idiotic.
We should now be able to compute the asymptotic limit of web-server bandwidth for slashdot-proofness per year for 600 years. I bet it's a constant price in street dollars.
I was a physics undergrad at Case, and actually had Starkman as a professor for a mathematical physics course. I have chatted with Krauss a few times since graduation on science topics involving public education. These are good guys, glad to see them headlining slashdot this morning.
"and that consciousness must be finite."
So they are saying that, using fundamental physics and mathematics, they have proof that if somebody has infinite wisdon, the universe can not be expanding?
--- Hindsight is 20/20, but walking backwards is not the answer.
This article contains a very large number of assumptions, which may well prove not to be the case (constant cosmological constant, no FTL communication/travel, no access to other universes etc. etc.). Still, an interesting intellectual exercise I suppose... ;-)
Galileo: "The Earth revolves around the Sun!"
Score: -1 100% Flamebait
Penrose is a mathematician who attempts to be philospohical and fails miserably, because he can't distinguish his intuition from fact. You don't need a link. Just remember that he wrote "The Emperor's New Mind", and coil away in horror.
Please use the mirrors. In Australia, the closest one is here.
Moore's Law is not a physical theory, it is the observation of a common phenomenon, namely the curve that technology goes through as it becomes cheaper and eventually free.
All technologies seem to obey this general law. Software, chips, disk space, they all tend to zero.
Even a passenger jet costs a fraction of what it did 20 years ago.
Moore's Law turns this around to say that for the same price we can expect more and more capacity. Long before 600 years are passed, this capacity will effectively reach "infinite", being the point where no-one can use more capacity or power, no matter what the application. At which point Moore's Law will gently slow down.
Ceci n'est pas une signature
"...consciousness must be finite."
This assumes that consciousness is based solely on computation. Not proven yet.
And for that matter, even if consciousness is nothing more than computation, how can we put a limit on an activity in space-time when we don't even know how space-time functions, or even how many dimensions it has?
Weaselmancer
Weaselmancer
rediculous.
Most likely, consciousness is a sense, like sight or sound.
Or, even more likely, an emergent byproduct of highly complex strange loops and pattern matching that, unlike any sense, does not have an explicit biological presence.
I haven't seen anything by Penrose which is like this. In fact, this article states an assumption ("consciousness is fundamentally computational in nature") that directly contradicts Penrose's most well known result, a rather dubious pseudo-mathematical "proof" that consciousness _cannot_ be computational as a consequence of Godel's Incompleteness Theorem.
So, no, it isn't really like Penrose's work.
There are some intersting ideas as to what the end of the universe could be.
There's also another theory about that if a couple particles collide with enough energy they can create a more perfect vacuum that would essentially "take over" the current universe (I suppose like an implosion). Maybe somebody knows the link for this.
I mention this as a backdrop for an interesting short story by Isaac Asimov called The Last Question. This link is a summary and contains significant spoilers, you may want to read the story first I think that it is apropos, as it deals with a powerful computer called Multivac.
This story is interesting to read, and interesting humanistic view. Good for pondering this slashdot thread/story. Good science fiction is useful.
Take the cheese to sickbay, the doctor should see it as soon as possible - B'Elanna Torres, "Learning Curve"
Not commenting on the paper itself, but it has been submitted to PRL, not accepted. It hasn't gone through that wonderful process of peer-review that is the very heart of the scientific method (that and falsifiability but thats another topic). NASA has been setting a particularily bad example here with science by "press release". PRL is not an easy journal to publish in, lets wait until other experts have a look and not cheat the scientific method like this. PRL should not be mentioned in connection with this paper until this get published - Anyone can submit a paper to PRL...
If there's only a finite amount of computation available, surely it's irresponsible to run things like SETI and the distributed.net cracking contests?! You're using up all of the sums, dammit!
Ydco co
"I guess the main argumentation of Roger Penrose is that Godel Incompleteness Theorem can not be understood by a computer,"
Penrose is just a carbon chauvinist with a chip on his shoulder. I've never seen him once offer actual proof of any such conjecture, only carbonist assertions that he can magically understand something that silicon-based life cannot: our future silicon overlords have a special place in virtual Hell reserved for his uploaded consciousness.
They claim that every computation step requires at minimum energy of ln 2 k_B T (k_B is Boltzmann's constant, T is the temperature of the system). This is only true for irreversible operations such as setting or erasing a bit.
But computation doesn't have to be irreversible. There are various proposals on how to build reversible computers that don't consume this minimum energy per operation. More information about reversible computing can be found in this introduction.
OS Reviews: Free and Open Source Software
Here's Turing's counter-example to bust Penrose's theorem:
If a machine (human in this case) can simulate a single Turing machine, and a Turing machine can simulate it, then it is exactly as capable (though perhaps not as quick) as any other Turing machine.
The first part is easy to prove: Any student who has learned Automata Theory should be able to simulate a Turing machine in their head, though it will be VERY slow and tedious.
The second is harder, but there is no reason to think that a simulation of every particle that makes up a human, plus a small environment (air, ground, food, water) around her/him will successfully simulate consciousness. The fact that today's computers are not strong enough doesn't invalidate humans being bound to a Turing machine's capabilities.
Any Turing machine is computational, therefore if the applications of Turing's thesis to humans holds, humans, and every part of them, including consciousness, are computational.
As far as Heisenburg's uncertainty theorem and quantum mechanics goes, it can be inserted into the simulator using rand().
Godel's Incompleteness Theorem doesn't apply to Turing's Theorem. Godel is talking about that there exists inconsistencies in any sufficiently complex langage (ie., the statement "this statement is a lie."). It doesn't contradict Turing's Theorem, since to disprove Turing's Theorem, we'd need to find a Turing machine that is incapable of simulating another Turing machine. All Godel says is that there will be non-sensical or impossible states in any Turing machine, but the machine can still work. (the proof that they exist is that English syntax can be programmed into any Turing machine, and the "this statement is a lie." statement inputted into the machine).
And as far a philosophy goes, so what if I'm limited to 2^2^40 states. I'll never get anywhere near experiencing all of them in the life of the universe, assuming I live that long. And in the same way that computers can execute computer games with fantasy themes, a computation human has nothing interfering with dreaming, pretending, or religion (though it might point out the silliness of latter).
Ummm.... no. Godel proved that the axiomatic system of Russel's PM allows the construction statements which can neccessarily neither be proven true nor proven false. There are other axiomatic systems that can be complete and consistent; IIRC it was in fact Godel who proved that the first-order propositional calculus is complete and fully consistent. Godel's fork only attaches to systems that allow the construction of statements about statements; many propositional systems (like the first-order propositional calculus) do not.
Oy.... where to start? Kant's Critique of Pure Reason is nothing but 600 pages describing how people distinguish intuition from fact (though admittedly Kant was using "intuition" in a sense that we don't normally use it today). Descartes wrote his Meditations as an attempt to remove "intuition" (again, closer to Kant's sense of the word than ours, but still) from philosophy. Plato, of course, says nothing about the subject directly but narrates several dialectical processes about the subject.
All's true that is mistrusted
Well... in 600 years of Moore's Law, you get 400 doublings of computing power. So you need 400 more bits in your key space than you think you do now. :-)
[TMB]
Penrose in recent years isn't saying "consciousness isn't a computer." Rather in collaboration with Stuart Hameroff and a number of physicists is saying that "consciousness is a quantum computer."
/.'ers whose first reaction is: "He says we're not computers. Uncool!" consider the contrary reaction: "He says we're quantum computers. Way cool!" Also note that, as all /.'ers should know, quantum computers don't have the same limitations as conventional computers on capacity, thus the well-known threat they pose to encryption, being able to break it (in theory) in trivially short time periods.
So for all you
"with their freedom lost all virtue lose" - Milton
I haven't read SotM, so I can't comment on that. My biggest problem with Penrose is that I personally am convinced that strong AI can be achieved, while Penrose believes that it is impossible. That does not mean that the book can't be good, but Penrose doesn't do a good job matching his arguments with those of his opponents.
He has a tendency to repeat arguments that strong AI supporters give, but in a way that they can be misunderstood, and then he misunderstands them and basically says they are stupid. Already in chapter one of ENM he starts ridiculing strong-AI supporters, without giving arguments. Later those arguments follow, but they are seriously flawed.
For example, while discussing Searle's Chinese Room experiment, he suggests that strong-AI supporters believe that "understanding Chinese" is in the book, and argues that it is stupid that a book can "understand" anything. But Turing's view (which is repeated by many others) is that the understanding is not in the book, but in the book + the human that reads the book. There is an emergent understanding of Chinese that comes from a book that describes how Chinese can be perfectly translated, and a human who strictly follows the rules that are written down in the book. This argument is mentioned by Penrose somewhere, but he just puts it aside as a very weak argument, that does not need refuting.
Writing like this annoys me immensly, and I suspect most my colleague AI researchers.
Incidentally, I think Penrose' work in mathematics is absolutely brilliant.