Scott Aaronson argues that the Singularity is far from here. He makes many good points. The one that intrigued me, however, is the mention of humans thinking much faster than we do:
fter all, we might similarly expect that there should be models of computation as far beyond Turing machines as Turing machines are beyond finite automata. But in the latter case, we know the intuition is mistaken. There is a ceiling to computational expressive power. Get up to a certain threshold, and every machine can simulate every other one, albeit some slower and others faster. Now, it’s clear that a human who thought at ten thousand times our clock rate would be a pretty impressive fellow. But if that’s what we’re talking about, then we don’t mean a point beyond which history completely transcends us, but “merely” a point beyond which we could only understand history by playing it in extreme slow motion.
There are two points of interest here. First is that you can’t do better than a Turing machine; all you can do is build a faster one. It is not clear to me whether the human brain is a Turing machine or not, although intuitively I would expect that it is. (Otherwise you have to propose some sort of magic arising either from, well, magic, or else from massive parallelism – and the latter should be emulable with sufficient computing power.) If that’s true, then there is some upper bound on the amount of intelligence you can cram into a human-sized skull. Of course, you have to be careful how you define ‘intelligence’. You might take out all the useless other-human-predicting modules and cram in some extra mathematics and logic instead, Asperger-style. Or vice-versa, for the ultimate politicians. Or you could do something a bit different, like say conscious control over the heartbeat (or other autonomic functions), much as we have over breathing. The point is, there is some upper limit to computational power contained in a brain. It is at least conceivable – I don’t think it likely, but it’s possible – that the most intelligent members of our species are already as smart as it is possible to be without fundamental structural changes to the brain. (In most cases, of course, they are smart in exactly one dimension, although there is at least some cross-over – whatever it is we call ‘g’, general-factor intelligence, does seem to be applicable to many domains. Real idiot-savants are quite rare.) If that’s true, then in some sense it’s a cheering thought – it means that, while we may have to worry about robots or AIs taking over, we won’t see genetically engineered superhumans doing so, at least not until we get good enough to make those fundamental structural changes.
The other point is the idea of speeding up a human by a factor 10000. I don’t think Aaronson has thought this through. Consider: I can move my hand across my field of vision in roughly a second, just moving it with no particular exertion. If I were speeded up by a factor 10k, that would seem to me to take almost three hours! Ye gods, the boredom! By today’s standards I’m fairly intelligent, with a lot of mental resources; I can usually find something to occupy my mind when I’ve got nothing else to do. But there are limits! It takes me about a second to flip the pages of a book; a three-hour wait between every page? Aaargh! And what’s worse, it only takes me maybe a minute to read the page I just flipped to – presumably it would be the same subjectively. So that’s a 3/500 ratio of reading-time to waiting time! (Actually, you might even run into a limit where you comprehend the page much faster than you can move your eyes. So even the actual reading… would… seem… like… ohgodthisissoboring waiting… forever… for… a… fantastically… slow… streaming… text… to… willyouforgodssake move.)
Now, I can move my hand faster if I try, and you can give me stronger muscles and whatnot, eyes with a larger focus area, books with pages that flip automatically and much faster, or perhaps a reader that scrolls very rapidly. But the fundamental problem remains: A day contains only just so many events. (And, by the way, if you think exercising is boring, just try it when it takes 10k times as long for the same result…) Oh, and speech! You can’t speed up speech very much, our tongues will only move so fast. Same for typing at a keyboard, or whatever input device is suitable: Your fingers only go so fast, no matter what you do with stronger muscles. (Note: I can move my hands a meter or so in a second with a bit of push. Multiply that by 10k, and I’m causing sonic booms. Or, more likely, breaking my fingers on the air resistance.)
So, our 10k-human has only a very occasional external event to react to. Even talking to other people as fast as him is an exercise in frustration, because they may have the processing speed, but they don’t have the I/O capacity. (Short of direct mind-computer interfaces, anyway. That might work.) Just walking across the room takes him several hours, subjective time! What’s he going to think about while brushing his teeth? Even top-level geniuses can’t be thinking brilliant thoughts all the live-long day, especially when the day is roughly 20 years long. (Counting only waking hours, that is. Which is another problem, actually – does he need to nap every 6 seconds or so, for three seconds each time?) To occupy himself he’s going to have to connect his brain to his computer and spend his time reading other people’s blog posts, on whatever subjects 10k-humans find interesting. (Long essays about the boredom of tooth-brushing, perhaps, composed at lightning speed. Or poetry. “Ode to the hummingbird: Little bird, why so slow? Move your wings! Life is brief, you haven’t got time to dawdle!”) And let’s note, at 10k times our current speed, download speed actually becomes noticeable. I can remember 3000-baud modems; the best modern connections would feel a bit like that, only without the crisp zippyness and zing.
I think, when the option of speeding up becomes available, that I’ll go for the basic model. A factor 2, let’s say. Greed, in this case, would be its own punishment.