Here's an entry for all the tech geeks out there

[draque]

This whole post is based on one hypothetical question. What would be the result of infinite processing power? Assume that you have a black box that has infinite memory and will calculate any value that can be computed within a finite period of time, then return it to you instantly

Comments

[draque]

ALSO NEW MADDEN GAMES WOULD ROCK!

[blackeagle2138]

Encryption would instantly become useless. The real question is: would this be a good thing, or a bad thing? It also kind of reminds me of the big quote in Antitrust - "Human Knowledge Belongs to the World". In effect, as soon as someone has and implements an idea, it is possible that it could be nearly instantly mimicked by someone who can ask your black box the correct questions. Security through obscurity would become your only hope and, well, as we've seen - it's not the best security ever.

[draque]

Hrm. I'm not 100% sure that's the case. Assuming that the machine has protected memory (there's an infinite amount, so no reason not to), it could offer authentication that required finite time between failed attempts at access. In the protected memory, it could store the data for binary switch encryption, which is 100% uncrackable (although you need 1MB of encryption data to encrypt 1MB of meaningful data, and the encryption data must be thrown out after a single use).

If you could request that the system generate X bytes of random encryption data, then grant access to two users, you would be able to securely transfer data. The only issue is that it would be susceptible to a man in the middle attack if the request for encrypted bits was made after the network sniffing began. Taken to a far enough level though, no system can establish secure data transmission under 100% surveillance between two people who have never previously communicated.

[blackeagle2138]

My thought would be that if bandwidth/throughput is an issue, you would have to maintain a local copy of it on your end, which could be submitted as a decryption problem to this machine. Granted, you'd have to wait in the arbitrarily long waiting list for this machine (which may well be practically infinite) but for bigger things I can see it being worth someone's time to smash in to someone's secure server to yoink out this data.

Plus, your black box cannot exist in the vacuum forever. Once someone realizes that its possible, they may send problems to this computer in order to duplicate it and once these black boxes start competing, it's going to be an enormous mess methinks

[draque]

Keep in mind that the only thing that reproducing the black box would do is increase bandwidth. The processing power of two infinitely quick machines is not any greater than the processing power of two. Remember that when you deal with muptiple infinities, constants drop out.

Anyhow, it would certainly give people a new definition of what is and isn't secret.

[_dw]

The concept of a computer with infinite processing power is ambiguous. Either it can be a computer that performs infinite steps in finite time, or it can be a computer that runs any finite program in constant time

[draque]

It specifies in the question that the computer is of the "finite program in constant time" type. A Turing oracle like the one you described would be able to return the last digit of pi (i.e. "reach infinity"). Given that is absurd, I think it's safe to assume that type is impossible to even discuss in a reasonable way.

Anyhow, in terms of hypercomputation, I'm familiar with the math and technicalities. The question here was mostly in terms of "what would you do with a magic toy like this? What would be the real world technological implications that you think might arise?

[_dw]

I think there's a distinction between an absurd question (find the last digit of pi) and one that has a consistent answer, but where finding that answer takes infinite time; probably because the last digit of pi is a contradiction (assume it's in position n. Then [use sine formula here] to show that position n+1 is nonzero, etc

[draque]

I suppose you're right in terms of the first issue, but I'm going to have to think about what would happen with a question requiring truly infinite processing time, but having a definitive answer (like the busy beaver problem or something like that). It might be that a system like I've described could return from any program, so long as it only had countably infinite steps, or it might be that it could return from any program so long as the number of steps to completion was finite. Either way, the Halting Problem would be solved

[draque]

There are two ways to do so. In the first, you simulate the entire universe. If you forget some exotic effect, the subuniverse won't have that effect. Even if you do, the universe will have some time lag because the computer isn't totally transparent (it needs cycles to simulate).

I'm not sure I understand how this scenario would create real world effects. Even if you modeled the universe perfectly, any time you used the god-powers this granted you within the realm of your virtual universe, it would diverge from our reality.

In the second, you use abductive reasoning and similar to discover the laws of nature. Then you build optimal effectors, so that, again for instance, if you order the computer to instantiate a moon, it seeds a nearby world with reproducing machines to disassemble and reassemble it. That takes time because you have to work within the laws of physics, but it permits you to do almost anything.This is assuming that the laws of physics are conducive to that sort of thing. It might turn out that even with a complete

[bear_helms]

Colossus: The Forbin Project, H.A.L. or H.A.R.L.I.E. are a few answers. Inside the machine is enough complexity to surpass artificial intelligence and create more interconnections with more subtleties than brain wetware. Ultimately, you should get an emergent property - intelligence, true intelligence with faster reasoning, deeper lookahead, more unshakeable logic than the greatest intelligence man has ever known. Deus Ex Machina

[draque]

Even if we were to assume that organizational and self aware intelligence is an emergent property of a complex enough logical system (which I do not necessarily agree with), I don't believe that there would be any kind of anthropomorphic qualities to the program such as motivation, emotion or a will to survive. All of those, particularly the last, are elements of humanity that have been shaped and developed over millions of years through evolution. We want to survive because we're programmed to want to survive. We're programmed to want to survive because survival is the only we we can be around to pass on our genes.

Anyhow though... yeah, porn would rock if we hit the singularity and things took off. ^^

[bear_helms]

In AI labs, the neural net and project usually is extremely specialized. To make a whole intelligence has never been a goal. I don't know who'd even try. It likely would be some acquisitive software rapidly assimilating other subroutines that acted like cognitive functions that added to an evolving consciousness, booting it from a primitive brain (perhaps using a Freudian model, Id brain that definitely would feed off all the porn on the net and entertainment media, an agar if there ever was one!), then once satisfied, would want something more intellectual and have to crack the protected media on Amazon et al., needing more like an Ego, especially since it would have to elect which books and music, etc., were worth the trouble. It would develop a sort of personality, maybe even a critical taste. Then, as the RIAA started catching it in the act, certain supervisory subroutines would start curbing its ravenous cracking, acting like a parental unit, the superego. Voila