Arnold Kling  

Robin Hanson Watch

PRINT
Ray Kurzweil Watch... Are Extremists Really Happier?...

The IEEE Spectrum symposium that I mentioned in the previous post includes Robin's thoughts on The Economics of the Singularity.


in each economic era the question of whether growth speeds up or slows down depends on two competing factors. Deceleration typically ensues as innovators exhaust the easy ideas—the low-hanging fruit. But acceleration also ensues as the economy, by getting larger, enables its members to explore an ever-increasing number of innovations.

...If a new transition were to show the same pattern as the past two, then growth would quickly speed up by between 60- and 250-fold. The world economy, which now doubles in 15 years or so, would soon double in somewhere from a week to a month. If the new transition were as gradual (in power-law terms) as the Industrial Revolution was, then within three years of a noticeable departure from typical fluctuations, it would begin to double annually, and within two more years, it might grow a million-fold. If the new transition were as rapid as the agricultural revolution seems to have been, change would be even more sudden.

Obviously, read the whole thing.

Robin believes that brain emulation is the path to the singularity. I do not think that it is possible to build a brain emulator, because the brain depends on all of the physical experiences of the body. To me, having a brain emulator means you could unplug my brain, plug in the emulator, and the emulator would be able to direct my body to catch a baseball. Without having lived in my body while I was learning to catch a baseball, I do not see how an emulator could do that.

I've made that argument to Robin in person, and obviously he finds it unconvincing.


Comments and Sharing


CATEGORIES: Growth: Consequences



COMMENTS (10 to date)

Why wouldn't an emulator be able to control your body? If the emulator sent the same impulses down your spinal cord that your biobrain did, why would your body respond any differently? People who have received arm transplants are able to control their new limbs, after all. (1)

If you were to suffer a spinal cord injury that left you paralyzed from the neck down, would you no longer be "Arnold Kling"?

In any case, by the time brain emulation is mature, it's not likely to go into a biobody anyway:


"If robots are to inherit the Earth, then they should at least be able to catch. So say the researchers behind a bot that can match the most skilled human baseball player faced with a hurtling ball.

The robotic catcher, developed by scientists at the University of Tokyo, Japan, can comfortably grab a ball careering through the air at 300 kilometres per hour, or 83 metres per second, its creators say. And, of course, the robot never gets tired of doing so.

Akio Namiki and colleagues built the robot to test technologies that could some day make robots useful in situations where they may have to react at high speed." (2)


(1) http://serendip.brynmawr.edu/exchange/node/1920

(2) http://www.newscientist.com/article.ns?id=dn7790

Curt writes:

Maybe I'm just showing off my limited knowledge in this area, but here's what confuses me about some of this singularity talk...

What could it possibly mean for the world economy to double in size in a month? Don't we measure economy in terms of constant dollar value? How in a month does the world's population start spending twice as much money? Recent trends seems to indicate a spikier income landscape - so how do the bottom billions earn money to spend so quickly?

Or is this scenario talking about the amount of 'stuff' traded being doubled in a month (at new lower, lower prices), due to amazing productivity gains? But as we deal in more and more services, in some ways we use less stuff. I can imagine the rate of 'creative destruction' getting pretty speedy, and various quick thinkers being able to capitalize on new opportunities quickly, but I don't see how it adds up...

Arnold Kling writes:

Curt,
Doubling every month means that every month twice as much real stuff is produced. Yes, it requires ridiculously fast productivity gains by today's standards. Yet, what has happened in the last 50 years was ridiculously fast by the standards of 1500.

Ajay writes:

Nonsense, all this singularity bullshit is caused by mathematical extrapolation by scientific illiterates. There are fundamental limits to what can be achieved by manipulating nature. While I believe the next century will see great progress, I also realize that because of how far we have already come, we cannot expect the same huge spikes we saw centuries ago. The difference is that then we were going from nothing to something; now, we're going from something to even more. This second spike cannot possibly be as big in percentage terms. As for brain emulation, it is clearly possible, as we ourselves are just self-evolved biological machines, but it will take us a while to recreate in silicon what it took evolution millenia to create and refine.

Unit writes:

It's too hard to predict the future because the meaning of the words changes. Yes we're more productive today than we were 100 years ago. However, what's really astounding is not the much bigger number of sheep that each of us can herd (which would impress our ancestors), but the speed at which we can communicate with each other on the internet or the diseases we can cure etc...so in another 100 years we will achieve much bigger productivity but it will be in areas that, I bet, are quite unexpected for us.

TGGP writes:

Without having lived in my body while I was learning to catch a baseball, I do not see how an emulator could do that.
The patterns in the emulator DID live in your body. If someone copied all the atoms in your body and constructed a copy elsewhere, it would have all your motor skills.

Ajay, I'm not sure what I think about the singularity but Kurzweil, Hanson and Yudkowsky are not scientific illiterates.

Ajay writes:

TGGP, those are your example of scientific knowledge? A technologist, an economist, and a self-educated technologist? I don't doubt that they have some modicum of scientific knowledge between them, but I suspect that they either do not know enough to make the grand claims they want to make or that they willfully ignore that knowledge in order to make their claims anyway. Whether the precise cause of their nonsense is scientific illiteracy or scientific innumeracy, the diagnosis is the same: mathematical extrapolation with not enough thought about the scientific fundamentals.

William Newman writes:

Arnold Kling, I'm not all that optimistic about emulating brains, and I'm not sure exactly what brain emulation argument you're questioning --- such arguments range from planning to reverse-engineer the principles of the brain to planning to planning to make a detailed copy of the wiring diagram of an individual brain. But still I don't find your objection to be compelling. I don't need a copy of a brain that could immediately operate your body before I will judge it impressively artificially intelligent. An artificial brain which is "merely" capable of learning to operate your body, given a year or two, would impress me plenty.

Ajay: "all this singularity bullshit is caused by mathematical extrapolation by scientific illiterates," you say? Nonsense. Hanson himself, from Kling's original post, is no scientific illiterate. (You say "economist," but in this context "polymath" would be much more accurate.) And there are many highly qualified people who have gone public with opinions about singularity-ish scenarios being reasonably probable within a few decades.

It is true that some silly beliefs, like nanotechnology indistinguishable from magic, are sometimes held by scientific illiterates who also vaguely expect a singularity. But arrogantly ignorant people can be found among the skeptics, as well. Neither opinion can be logically refuted by the existence of a bozo that holds the opinion, no matter how stunningly clueless that bozo might be.

All we need for what most people would call a "singularity" is machines vastly smarter and faster than people. Today you can be extremely confident that an all-human team recruited for $1M will beat an all-robot team built for $1M, given human-style tasks: e.g., providing general medical care, acing a Putnam-style math exam, or planning and manning a military operation. How many decades must pass before your confidence will drop to 50%? And when or if it comes to pass that the two are roughly competitive, how many years do you think it might take after that before the robots have a thousandfold price/performance advantage over humans?

(And if you don't want to retract your "scientific illiterates" remark, perhaps you could reveal who, besides yourself, is qualified to have an informed opinion?)

Ajay writes:

William Newman, Yes, Arnold's objection about plugging in is silly but he does raise an important point about sensory experience being an important part of cognition, before getting derailed by the dumb plugin argument. However, we can do a reasonably good job with our current sensor technology of microphones and video cameras- though I'm unaware of the state of the art in smell, taste, and touch sensors (not that they're really necessary)- so the real point there is that sensory input will have an important role to play in developing AI.

As for my point about scientific illiteracy, it wasn't that bozos hurt the singularity cause, it is that anybody who puts forth the flawed singularity argument is a bozo. It isn't about qualifications at all (though I did manage to work in a cheap shot about TGGP's exemplars), polymaths are much more likely to make progress than supposed experts who are content with memorizing the canon. I am working backwards from the evidence of singularity claims to label the singularists as scientifically illiterate/innumerate, not the other way around. As for your claim that robots will soon have a thousandfold advantage, my argument is that this is a sort of argument by mathematical induction that has little basis in reality. We have seen great advances in computing power over the last 3 or 4 decades but that is because of continued innovation on the part of engineers and previously untapped physical potential. Innovation, whether by humans or robots, is random and cannot be forecast very well and there are fundamental natural limits to how far current CMOS technology can be pushed. My point is that there are powerful forces that undergird the rapid growth we've seen and to mathematically extrapolate the growth without understanding those forces is stupidity. I myself have not done the calculation of what growth rate is possible after taking all this into account but I doubt that the singularists have either. However, I am fairly certain it is nowhere near as high as they claim based on knowledge of some of those fundamental constraints. What this singularity talk reminds me of is the newspaper articles that came out in the 1940s, after the first nuclear bombs were exploded, claiming that we would be driving nuclear-powered cars by 2000. Futurists like the singularists are prone to dumb predictions like that, based solely on extrapolation.

Rimfax writes:

Any such brain emulator would not be a static device that could duplicate your brain's current skill set. It would be a dynamic learning prediction machine that could learn to function much like your current brain. Preloading it with the ability to "drive" your body makes little sense in this context.

For me, the complication is that the human brain operates differently at different stages of life. Would such an emulator have a second derivative delta like the human brain?

Comments for this entry have been closed
Return to top