Slashdot Mirror

http://www.jetpress.org/volume...
"This paper describes how the performance of AI machines tends to improve at the same pace that AI researchers get access to faster hardware. The processing power and memory capacity necessary to match general intellectual performance of the human brain are estimated. Based on extrapolation of past trends and on examination of technologies under development, it is predicted that the required hardware will be available in cheap machines in the 2020s. ... It may seem rash to expect fully intelligent machines in a few decades, when the computers have barely matched insect mentality in a half-century of development. Indeed, for that reason, many long-time artificial intelligence researchers scoff at the suggestion, and offer a few centuries as a more believable period. But there are very good reasons [exponential growth] why things will go much faster in the next fifty years than they have in the last fifty. ..."

This reminds me of some speculation on the rate of innovation in reaction to the attention Neil Gershenfeld's Fab Labs have been getting.

Fab Labs are USD$20,000 boxes of equipment that can be used, with an afternoon of training, to cook up many inventions, including basic electronics. This is the beginning of something that eventually might resemble a desktop nanofactory like this one proposed by Chris Phoenix of The Center for Responsible Nanotechnology.

In short, the premise (open for debate, at present) is that personal fabricators ("PFs") in every home could do for patents what personal computers (PCs) in every home have done for copyright, when combined with the Internet. If the tools to innovate material things were distributed to non-commerical users, the amount of non-commercial innovation increases, whether we're talking about software or gizmos. For twenty years, incumbent comemrcial patents might dominate, but once they expire, openly-licensed patents could dominate through sheer numbers.

The popularity of open source licensing has largely solved the copyright problem; there are plentiful alternatives for those who insist on software freedom, and more are arriving daily. PFs could extend this to physical inventions, as well.

Huebner's choice of "major innovations" and patents as his measures of innovation just prove his bias is slanted toward the fact that innovation is decreasing because it's concentrating along with the wealth that results from it.

10 to 15 years? :)

When biology makes stuff, it needs to make it with protein, which isn't a very stiff/sturdy material. So living things are soft and mushy (even Ahnold) compared to, say, building materials for skyscrapers or airplanes. Also, it's tough to build good computers out of biological materials, and without a couple billion years of evolution, we'd be as dumb as lettuce.

So if nanotech can make smarter artifacts out of stiffer materials than biology can, it's conceivable that nanotech could overrun biological life. IIRC, Drexler's original horror story was that the _entire_biosphere_ could be consumed; maybe if the experiment were really done, we might find some equilibrium where some biological life still existed.

It turns out that gray goo is a very tough engineering problem. Imagine building a car that could wander through the woods, foraging for fuel. The energy density in the woods is very low on average, and the car is unlikely to find the energy it wants before it runs out of its previous fuel supply. The varieties of fuel-like things that it might find is staggering: maple sap, starchy vegetables growing wild, beehives with honey in them, dried wood, a half-empty gas can in some barn, or the energy in the food in a kitchen. If the wild car is to be a real analogy for gray goo, it must recognize each possible energy source and be able to use it. Hard problem.

Because it's such a hard problem, nobody will do it by accident, any more than somebody will "accidentally" land safely on Mars. Also, whatever you're trying to accomplish by making gray goo, there is almost certainly a simpler cheaper way to accomplish it. Want to wipe out humanity? Make an AIDS variant that's mosquito-borne. Or do the same with ebola or smallpox. A few people want to wipe out humanity, but far fewer people want to wipe out all plant and animal life on Earth.

The point of Drexler and Phoenix's recent work, and this probably isn't quite emphasized strongly enough in these recent commentaries, is that gray goo isn't even a very good manufacturing paradigm. Chris Phoenix has been thinking a lot recently about sensible nanotech manufacturing strategies, and they look a lot more like SIMD vector processors. You have a broadcast instruction stream driving a bunch of individually stupid actuators. Duplicating the program store and instruction decoding for each actuator is wasteful, in terms of both material and reliability.

One reason for this is if it takes you 3 or 5 billion years to get all the information back from many light years away there may be a significant possibility that you or the message transports will be exposed to a Gamma Ray Burst that wipes a significant fraction memory. The second reason is that you can't have a significant "conversation" across millions of light years -- the chances are significant that you will have computed the answer yourself by the time you send a probe a million light years away and get a response back from some colonized system. Finally we have know for the last ~25 years (since the Project Daedalus study that you can't send probes out, matter based messages back, etc. at anything close to the speed of light. The energy released when you strike a piece of space dust or even particles from interstellar solar winds is going to erode the ship. Proj. Daedalus solved this by building a big shield on the front of the ship. But if you do that you are going to need a lot more energy to accelerate ships or message transports to higher velocities. Many of these problems are discussed in the Matrioshka Brain papers.

I would suggest that Krauss & Starkman may be very good physicsts & astronomers but I would think twice about putting them on an engineering project.

So the question must be raised *what* would the Universe look like if /.ers had had a billion or more years to work on it? Yes, I know that many of you will argue that it should not look much different but you have not run the numbers as I have on planetary disassembly times. Nor do you understand the limits of nanotechnology to the extent that I do.

I've tried to explore and address some of these questions in my papers about Matrioshka Brains as has Dr. Sandberg in his exploration of the various types of Jupiter Brains.

These are not new concepts -- they have been discussed on the Extropians list for perhaps a decade. There are a few good astronmers and astrophysicists who discuss these ideas but to a large extent mainstream science seems stuck in the paradigm that the universe simply must be dead.

Until we deal with whether or not that is a fundamental misconception we may be plagued by concepts like Dark Matter and Dark Energy that could be resting on very questionable evidence.

Robert

It's my belief that this failure is largely a result of hardware and not software. Current analysis of the processing power of the human brain suggests that the raw computation power available to AI scientists have been pathetically inadequate. In the 80's researchers using PCs would have had the processing power of approximately a worm. There's only so much you can do with a small brain: tasks truly useful to humans (visual recognition, for example) are going to be impossible without similar processing abilities. There are good reasons that ants use chemical detection to identify other members of the colony instead of visualy recognizing them. This is starting to change.

It's all handwaving, the nano pundits can't put forward any kind of actual theoretical design for a universal contsructor.

Actually, several people have proposed designs for nano factories (and thus universal constructors). Incremental progress towards the components of these nanofactories is being made every day.

It's my belief that this failure is largely a result of hardware and not software. Current analysis of the processing power of the human brain suggests that the raw computation power available to AI scientists have been pathetically inadequate. In the 80's researchers using PCs would have had the processing power of approximately a worm. There's only so much you can do with a small brain: tasks truly useful to humans (visual recognition, for example) are going to be impossible without similar processing abilities. There are good reasons that ants use chemical detection to identify other members of the colony instead of visualy recognizing them. This is starting to change.

It's all handwaving, the nano pundits can't put forward any kind of actual theoretical design for a universal contsructor.

Actually, several people have proposed designs for nano factories (and thus universal constructors). Incremental progress towards the components of these nanofactories is being made every day.

I'm not sure on the accuracy but this figure seems to indicate that long before Moore came along (and the transistor, of course) computational power was increasing at a roughly exponential pace since the introduction of the electromechanical calculator. The figure is taken from a larger paper discussing the computational power of the human brain.

I'm not sure on the accuracy but this figure seems to indicate that long before Moore came along (and the transistor, of course) computational power was increasing at a roughly exponential pace since the introduction of the electromechanical calculator. The figure is taken from a larger paper discussing the computational power of the human brain.

However, the promotion of SETI@Home by anyone demonstrates they have not looked at the problem in detail.

There is reasonably extensive documentation on the probable intelligence of advanced civilizations (for example see papers by Dr. Anders Sandberg (here) or myself (here). As I have pointed out at conferences and in papers the difference between an advanced civilization and the human civilization is ~10^24 Ops. The difference between a single human and and a nematode worm is ~10^15 Ops. We don't talk to worms and advanced civilizations don't talk to us!

Furthermore the entire SETI effort does not take into account the information content of an advanced civilization. By my estimates this is of the order of 10^50 bits (probably more). One cannot communicate even an extremely small fraction of that information content across interstellar space using radio waves. They simply lack the information carrying capacity. So the SETI Institute, Drake, Tarter, Shostak, et al have sold millions of computer users (as well as Paul Allen) a "bill of goods" without having done their fundamental homework on the limits of evolution of civilizations. Why on earth would one attempt to communicate with a civilization that is fundamentally less sophisticated than a nematode worm and with whom it is impossible to exchange a significant amount of information that one has at ones disposal?

In contrast Marvin Minsky (probably one of the leading AI experts in the world) and Freeman Dyson (a brilliant mathematician/physicist who should have won a Nobel Prize for his contribution to the Tomonaga/Schwinger/Feynman contribution to quantum electrodynamics were it not for the Prize limits of 3 individuals) had this worked out in 1971 at the conference between Russian and foreign scientists at the Byurakan Astrophysical Observatory. Direct quote from the proceedings edited by Sagan:

MINSKY: Since radiation at any temperature above 3 deg. K is wasteful and a squandering of natural resources, the higher the civilization, the lower the infrared radiation. We should look for extended sources of 4 deg. K radiation. There should be very few natural such sources.

DYSON: I don't quite go along with this but to some extent you are right.

Minsky obtaining a concession from Dyson is significant. It has been ignored by the "radio waves from aliens" camp. They *will not* be trying to talk to us. But we *might* be able to observe them in the IR detection region. (Unfortunately IR detection is difficult to do from ground based telescopes.)

So the bottom line -- reallocate your spare computer resources to projects like folding or in the future to Nano@Home. SETI@Home is never going to succeed. It is based on outdated fantasies. Telescopes like the failed WIRE mission or the recently launched SIRTF *may* be able to detect alien civilizations but efforts such as SETI@Home are pointless until such time as the supporters make the case that advanced civilizations would want to waste their time communicating with sub-worm civilizations.

Robert

Obviously we cannot now be sure that we are not living in a simulation. The more likely our descendants are to be rich, long-lasting, and interested in simulating us, the more simulations of people like us we should expect there to be on average, relative to real people like us. And so the more we expect our descendants to be rich like this, the more we should expect that we are in fact living in a simulation