[Continuing coverage of the 2009 Singularity Summit.]
Michael Nielsen is beginning his talk, “Quantum Computing: What It Is, What It Is Not, What We Have Yet to Learn.” (Abstract and bio available here.) He promises it will be a broad overview. I hope so after that last talk.
He’s talking about the dangers of the models we use to commonly talk about quantum computers — dangers, that is, in how misleading they can be. Now he’s talking about the model wherein a quantum computer is supposed to be one that can be in many different states in once. The problem with this is, he says, that if it were true, we could directly and efficiently simulate it on a conventional computer. But the interesting thing about quantum computers is supposed to be that they can do things that conventional ones can’t. He calls this the “quantum gap,” which results when we try to understand quantum computers with conventional models.
This should be really fascinating. My recent piece on A.I.
toward the end hinted at how much of the hope for the Singularity rests on whether quantum computing will allow us to get around the theoretical limitations in conventional computers (which is not to say current limits in speed or memory). Is there some end run around the problem of having to express intelligence as a computable function?
Nielsen is promising now to tell us the complete mathematics of quantum computing in less than ten minutes. I think it would be better for me to post the video of this portion rather than attempt to summarize it. [UPDATE: My video hasn’t turned out well; I’ll add links later when and if I find other video online.]
He’s finished with the technical stuff now, and is describing the known uses for quantum computers. There aren’t many, but there are some specialized uses, like factoring numbers far more quickly than in conventional computers (useful for breaking encryption systems). He says this is all “widely believed” — meaning of course that while it has been discussed theoretically, it has not yet been done in practice.
Now Nielsen is asking what quantum computing has to do with the Singularity. He’s voicing “one thought,” which is that many have speculated we don’t understand quantum mechanics because we evolved in a classical world. Maybe, he says, a quantum artificial intelligence will be able to understand quantum computers because it is made from them. He doesn’t explain at all how experience factors into all of this…
The other important point, he says, is that maybe the “first wave” of conventional AIs will be able to harness quantum computing, and usher in a second wave of even better quantum AIs. Maybe, he says, a few hours after the first Singularity, when we humans will already be unable to understand the new innovations that are occurring, the computers will themselves create another class of computers in another Singularity that they themselves will not understand. (That sounds pleasant.)
Questions now. Ray Kurzweil has the first. He says that was the clearest description of quantum computing he’s heard. (Hey, that’s not a question, Ray!) Kurzweil says he doesn’t think we need quantum computing to achieve Turing Test-level AI. As for whether we need it for consciousness (something someone asked during Nielsen’s talk), he says he’ll think about that during the day and get back to us during his own talk later. Uh huh.
Kurzweil asks another question about scalability. Nielsen says the progress now is slow, but that, like with the original computers, it won’t chug along linearly — we’ll find new methods as we go along. And that’s that.
October 3, 2009