January 02, 2007
Greg Cochran Says Replicators Will Solve Energy Problems

Physicist and genetic theorist Gregory Cochran thinks within decades we will have more energy than we can imagine and incredible wealth too.

Hardly anyone seems to realize it, but we're on the threshold of an era of unbelievable abundance. Within a generation—sooner if we want it enough—we will be able to make a self-replicating machine, first seriously suggested by John von Neumann.

Such a machine would absorb energy through solar cells, eat rock and use the energy and minerals to make copies of itself. Numbers would grow geometrically, and if we manage to design one with a reasonably short replication time—say six months—we could have trillions working for humanity in another generation.

Regarding getting technological advances sooner: The lack of a more vigorous pursuit of the big pay-off breakthroughts has got to be the absolutely hugest opportunity cost we inflict on ourselves. We've somehow managed to allow the United States to get in a position where it is going to waste about $150 billion in Iraq this fiscal year. Yet the total budget for the US National Science Foundation for a wide range of research efforts in many areas of science is about $6 billion dollars. Granted, other research agencies get much larger chunks of money compared to the NSF. But ridiculous fiascos get far more.

The vast bulk of the wealth gained from advances in knowledge go to people other than those who generate the knowledge. So marketplaces really under-reward those who push back the edges of scientific understanding. We therefore do not get as much net benefit from science as we could. We underfund science.

Self-replicating devices (which better have great built-in controls for stopping replication) can make solar photovoltaics cheap.

Right now the human race uses about 13 trillion watts: the solar cells required to produce that much power would take up less than a fifth of one percent of the Earth's land surface—remember that the Earth intercepts more solar energy in an hour than the human race uses in a year. That's a lot of solar cell acreage, but it's affordable as long as they make themselves. We could put them in deserts—in fact, they'd all fit inside the Rub' al Khali, the Empty Quarter of Saudi Arabia. As I understand it, we like depending on the Saudis for energy.

But there are better ways. Solar energy works better in space—sure, the weather is better, but also consider that the vast majority of the Sun's energy misses the Earth. In fact only about one part in two billion warms this planet. Space-based self-replicating systems could harvest some of that lost sunlight—enough to make possible a lot of energy-expensive projects that are currently impractical.

Greg also expects the incredibly greater ability to harness energy and matter to make interstellar probes possible within the lifetimes of some people who are still alive.

If Greg is correct then projections that we'll be burning huge amounts of fossil fuels in the latter part of the 21st century are entirely wrong. We could bring the age of fossil fuels to an end much sooner by pushing much harder to accelerate the rate of advance of nanotechnology. That'd pay orders of magnitude greater dividends than attempts to make everyone think warm fuzzy thoughts about hybrids and hostile thoughts about SUVs.

Share |      Randall Parker, 2007 January 02 11:08 PM  Nanotech Wealth

John S Bolton said at January 3, 2007 2:20 AM:

if solar collectors in space are in our 21st century future; so might there be constructed a mighty parasol, which could shade the places which get too hot.
Some places get too hot to rain; but if shaded, would stay cool and humid enough to be pleasant and lush.
Global warming, if it develops as a problem, could be still profitable to deflect.
Maybe aircraft can burn just the right fuel mix to generate contrails, and these can be layered over urban regions to cool them.

jb said at January 3, 2007 6:11 AM:

The problem with those reproductive controls, of course, is that replication errors will eventually remove them, and the unfettered robots will populate wildly, drowning out the well-behaved ones.

Now admittedly this is a problem that would probably take millions of years to evolve. :)

What we really need are two kinds of self-replicating robots - those that eat rock to reproduce (type A) and those that eat type A robots to reproduce.

In any case, self-replicating robots or not, there's no doubt in my mind that the near future has tremendous potential for uplifting the human condition.

Ivan said at January 3, 2007 6:43 AM:

The space-based generation of energy is a bit longer term than efficient solar.

Good nanotech will initially mean very smart materials. Self replication is more difficult.

If you have plastic photo-voltaics, they could be made to fill the deserts of the world -- as the quote describes.

To make that happen in space you need to account for the loss of energy in transmission, the energy loss in the atmosphere, the energy loss on reception, and the energy required to get the materials in orbit!

Also, good materials make batteries better -- so more infrastructure can be based on raw electric power.

Even if this doesn't happen for some time, you can have an engineered solution to global warming through space-based shade & solar.

Which would you rather see in response to the threat from global warming:
- A race to build a space elevator to build solar arrays
- A race to the bottom in de-industrialization

Ivan said at January 3, 2007 6:50 AM:

From here:

geostationary orbit (GEO) is approximately 35,786 km up. A perfect lift system would only require the potential energy of the list. For 1kg, that means 350,000,000 joules. How long would it take a kg solar array to get that much energy to a base station, including losses?

Garson Poole said at January 3, 2007 7:00 AM:

When constructing future scenarios it is essential to be aware of the concept of self-replicating machines and factories that use solar energy. However, it is helpful to realize that this concept has an historical lineage that stretches back for several decades.

One popularized version of the idea was presented to Discover magazine readers in 1995. The article entitled Robot, Build Thyself discussed a collection of robots that would replicate and fill parts of the southern New Mexico desert. The authors envisioned the construction of a huge array of solar panels that would feed electricity into the national power grid. They also suggested other tasks such as the desalination of water for agriculture.

The idea has other important precursors. In the 1980s NASA “proposed building self-growing mining modules on the surface of the moon.” In the 1970s “physicist Freeman Dyson conducted some famous thought experiments on the future of machinery. Among other ideas, Dyson proposed building a rock-eating automaton that would fill the Sonoran Desert with self-reproducing machines.”

The book “Kinematic Self-Replicating Machines” by Robert A. Freitas Jr. and Ralph C. Merkle is a remarkable survey and it is available online here.

Gerald Hibbs said at January 3, 2007 8:00 AM:

"That'd pay orders of magnitude greater dividends than attempts to make everyone think warm fuzzy thoughts about hybrids and hostile thoughts about SUVs." -- I take that as a personal shot! :-)

As interesting as the possibility of solving our energy problems with solar, I don't consider that the source of excitement for nanotechnology. We already have a potential source of essentially unlimited energy -- nuclear. What we don't have is a source of unlimted practically free Porsches. The possibility of having a home fab lab would end many of our social problems while creating all sorts of new existential dangers (think future versions of contemporary hackers except they work with genetics.)

Once you can live in the cheapest studio apartment (at this point land is the only physical scarcity remaining) yet have robot servants, the best quality electronic gadgets (probably through some kind of DRM for the design like modern legal song services), food based on the cooking of world famous chefs, etc then that removes a lot of the room between the haves and have nots.

While having space based access to solar energy would solve our energy problems it wouldn't drastically change human society and our way of life. But a home nanotech based fab lab would revolutionize the world in ways we can't even imagine.

I think the best way forward would be, rather than direct funding because who trusts the government to fund the right approach?, to establishing a road map of achievements that lead to working nano assemblers and having incrementally larger prizes for each step on the ladder. Were I a billionaire I'd fund it myself. It is unfortunate that guys like Gates and Soros spend their cash on relatively unimaginative goals of limited effectiveness and utility.

Kurt9 said at January 3, 2007 10:06 AM:

When I was younger and got caught up in the whole Drexlerian nanotech thing (late 80's), I believed that we would see this kind of self-replicating technology in another 20-30 years. 20 years has passed since then. I was also reminded of Robert Frieta's prediction in 1980 that we would be able to make a 100 ton self-replicating "seed" and send that to the moon in order to initiate the build-up of industrial space infrastructure by 2000. This prediction served as the basis of my favorite SF novel of the time, "Voyage from Yesteryear" by Jim Hogan.

Of course, none of this has come to pass. As time went on and I ended up an expat in Asia, I came to see this nano-geek vision of "replicator" technology as just a modern-day "cargo cult" wishfull thinking. Call me cynical, but I am wondering if "replicator" technology is similar to plasma fusion and artificial intelligence in that it will always be "30 years in the future", never to be realized in the present.

On the other hand, Gregory Cockran is someone whose opinions I usually respect.

I do not believe that "dry" (Drexlerian) nanotech is possible. Even if it is possible and is developed, I think its capabilities to be much more limited than its proponents postulate (this means the guys at CRN). "Wet" nanotech, which is often called synthetic biology or bio-nanotechnology, is feasible and could certainly lead to the kind of "replicators" that Gregory Cockran predicts. One useful indicator is that there are over 100 labs working independently of each other to develop synthetic biology or "wet" nanotech. These labs are distributed around the world in places ranging from Japan to Brazil. Given this much independent effort in this technology, maybe we really will see this kind of "replicator" technology in another 30 years or so. It would certainly make space settlement much easier and cheaper.

The website I usually refer to for info on "wet" nanotech is "softmachines" at www.softmachines.org, which is about the only nanotech website that I look at regularly.

Joe said at January 3, 2007 11:01 AM:


Who expected to have self-replicating nanotech in the '80s? We could not even image an atom in the 80s, know that CNTs existed, and computer power was millions of times more expensive. Just because prior predictions of an event didn't happen doesn't mean that the event wont ever happen. MIT, I believe, just put out a document of all of the different types of nano moters that have been created in the last few years. In the '80s did any research like that exist? Right now we are building the toolkit that will allow MNT to shine. I won’t give a timeline because that depends on non predictable events and variation in funding, which is at an all time low IMHO.

We need to advance on both the "dry" and "wet" nanotech fronts as both of them have different potentials.

K said at January 3, 2007 11:02 AM:

As Kurt9, and Garson, mention, the moon is the place to try such devices. Plenty of solar power and rock. The starter kit wouldn't be all that heavy and we should be able to get it up there.

100 tons seems a very high estimate for the 'seed' or starter kit, but that figure is from 1980. And if it is 100 tons we aren't going to do it for quite some time.

TTT said at January 3, 2007 12:18 PM:

I don't think the cost of the Iraq War ($350 Billion) is a waste *if* it results in a functioning democracy in that region of the world. This is a big if, however, and the jury is still out, and will be out for some time, on this matter. If it does work out in the long run, I believe the money was well spent.

Also, I'm not sure that the rate of technological change can be forced. The progress curves are quite smooth, and looking at Ray Kurzweil's work shows this.

For example, we put a man on the Moon in 1969 for political reasons. But today, the same mission would cost only about 1/20th as much, in proportion to US GDP, as it did in 1969. We spent so much so early for political, not technological reasons.

Brian Hayes said at January 3, 2007 12:35 PM:

I'm going social for this comment, looking at the fabric of development. As new ideas emerge in our busy world, it is not only scientists that are crushed. Hundreds of thousands of eager entrepreneurs are lost as well in an arrogant and libertine torrent of acquisition and disregard. Developers of ideas and invention face a challenging landscape that most of us would never dare to enter.

If we are lucky enough to alter our prospects for the future, I think we will need some changes to help reduce the hurdles that developers of new ideas are facing. There are so many injurious tricks in the trade of capital formation. Too many project developers with grand proposals will report how cruel it is to bring a product to market. It is so easy to lose your shirt.

Among dozens of caustic obstacles for innovators, it is too simple for investors to burn a patent by letting it lanquish only a few years until it's value rapidly diminishes. A cooperative pool should be created to distribute at least declining portions of parallel development rights. It is too easy to dominate contract rights. A 'business court' might be required where the bench is chartered to assure the cost of justice is allocated fairly. I've known victors of legal battles win their rights as they are poured into bankruptcy by unfair and dastardly tactics under the law. Plus research colleges must use better methods to secure revenue. Currently, too much research is paid by public money but sold to wealthy lurkers for a pittance. There might be new rules to assure that some portion of all income from regionally chartered research stay in the community and some portion attach to the researchers.

Wealth is a good thing. Our prosperity is a better thing.... Our institutional leadership fails us. While we achieve financial, technical and engineeering growth, we fail to innovate in our social institutions.

Bob Badour said at January 3, 2007 2:22 PM:

Brian Hayes,

I disagree. We have had far too much innovation in our social institutions with almost universally disastrous results.


Is it still a success if the functioning democracy habitually elects a cruel, bloodthirsty theocratic government that engages in wanton torture and oppression of minority groups?

Brian Wang said at January 3, 2007 2:35 PM:

Robert Freitas did not predict the 100 ton lunar seed. He made a fairly detailed plan as part of study of a self replicating system of robots. Nothing happened because no one who controlled large budgets in the government chose to try and make it happen.

Drexler wrote the books about Nanotechnology starting in the 1980s about what should be technically possible and showed with Nanosystems that the programmable control of molecules did not violate known laws of physics. Drexler proposed wet and dry systems. Richard Jones of softmachines likes to try to mark his territory and claim that the Drexler proposals were only for 'dry systems.' The dry systems had the advantage in the 1980s and early 90's of being easier to draw a diagram of to stick in a book. They were also systems that were more clearly describable. Like look here is a small seesaw, let me use it to explain how a lever and fulcrum system could work. Look at the small scale because of laws of scaling it can clearly be shown to have this high level of performance. It does not mean that the simplified and describable system is the optimal system to build.

How much money has been spent trying to directly make the goal of direct control of molecules happen for the "dry system" since 1980? Almost none. How much was spent to try to prove whether it could or could not be done? almost none. So if you do not really try to make something happen or to see it would work then why are you saying that it cannot be done ? the off the cuff, no money (and no detailed work) spent speculation that it cannot be done has not proved its case.

Robert Freitas and Ralph Merkle have performed computer simulations using computational chemistry with density functions. Their detailed work has shown that many steps of the dry diamondoid mechanosynthesis can be done. Dimers of carbon placed onto a surface can be done.

The Germanium version of the DBC6 dimer placement tool had detailed computer simulations performed showing that it should work reliably at room temperature. There is no problem placing dimers with a one dimer gap. Placing them directly adjacent can cause some defect formation.

The paper also describes using a first pass of placing every other dimer. Then adding the skipped dimers. This process is a defect-reduced procedure for fully populated dimer rows.

Some steps may require cooling to 80 degrees kelvin.

The point is that actually working the specifics of the problem can bring good results. Instead of saying well my gut feeling or the gut feeling of a PHD or nobel scientists who does not want to try to make it work and would rather fund his own projects on carbon nanotubes or 'synthetic biology' tells me that this will not work or that it will not have good performance. We could have asked GM and Ford about electric cars back in the 90s. Those companies could have bee considered car experts. They would have said that they tried with the Impact and it did not sell many units. But they did not consider hybrid systems and their interest was not in fuel efficiency but in selling higher margin trucks and SUVs.

DNA nanotechnology has made big progress recently.

TTT a redo of the Apollo missions will not be cheaper now. The new back to the moon NASA plans are more expensive than Apollo.

10 to 40 million per year now directly towards detailed work on diamondoid mechanosynthesis by guys like Ralph Merkle and Robert Freitas would have a big impact.

The indirect work on dual use things like synthetic biology might lead to programmable molecular control but it will take longer. Both should be worked upon.

Brian Wang said at January 3, 2007 2:38 PM:

What has been shown is that mostly unpaid volunteers have taken longer than 24 years to make molecular manufacturing happen.

TTT said at January 3, 2007 3:34 PM:

"Is it still a success if the functioning democracy habitually elects a cruel, bloodthirsty theocratic government that engages in wanton torture and oppression of minority groups?"

Bob Badour,

It is an unfortunate situation, but still is technically the will of the majority.

But why pose a hypothetical question when that is certainly not the case in Iraq today. Plus, if genocide bothers you so much, you should be very offended by what is happening in Darfur. Or Thailand. Or what once happened under Saddam.

Brian Wang,

The new missions are far more elaborate. As a percentage of GDP, a Moon mission of comparable scope would be 1/20th of the cost of what it was in 1969.

Ken said at January 3, 2007 9:01 PM:

I think we'll see new developments in manufacturing getting trialled and put into use here on Earth long before we see them sent to space. Be they Drexler style self replicating nanotech, Smalley style mass production of nanomaterials or improved automation of more conventional production systems, if they're successful they'll be put to use right here, without the added costs and difficulties of sending them to the moon or some asteroid, at least not in the near future. We need the improvements and we need to see R&D get funded, but adding a moon launch into the R&D efforts to develop low cost PV and superior energy storage and distribution systems seems to be choosing to put huge obstacles in the way. Sure there's real benefits to developing better space launch methods but I have to believe they ought not to be bundled with energy research.

Randall Parker said at January 3, 2007 9:18 PM:

John Bolton,

What we could do with space light reflectors is to reflect light away from major ice concentrations so that Greenland and the Antarctic ice do not melt. That'll keep sea levels low.

We could also reflect light toward crops early and late in the day to make days longer during growing seasons. Imagine Iowa and Nebraska with longer light days in September and October. Any time a frost threatened in April and May make more light reflect down to warm up the fields more.

At the same time, we could reflect light away frmo places that get too hot.

Climate engineering could boost crop outputs and prevent some of the worst effects of atmospheric CO2 increases. This'll become very cheap to do once we have nanotech devices that can do massive amounts of manufacturing.


Conversion efficiency will matter greatly for determining how fast space solar satellites can pay off. So will the thickness of the material required. We do not know how far photovoltaics materials research will push photovoltaics to make them lighter and more efficient. Also, nanotech beanstalks will lower the cost to putting stuff into orbit.


I picutre someone in the mid 19th century saying "When I was younger and caught up in the dream of creating heaver-than-air craft with wings that can fly". Past failures do not predict continued future failures.

Randall Parker said at January 3, 2007 9:33 PM:


One way to deal with the reproduction threat is to make several types of devices and have each type make a different type and not be able to make itself.

Another way to deal with the reproduction threat is to give them several techno inputs that we control that they are programmed to need to do reproduction. Make them need intervention in order to do a reproduction cycle. Don't let them run open loop.

Still, some fools will end up designing modifications that do not need help in order to reproduce. Look at all the fools who write computer viruses. Some do it just for glory.

Randall Parker said at January 3, 2007 9:41 PM:


I think money spent on developing better space launch vehicles is money wasted. We are better off spending the money to develop materials that can be used to make better everything and faster and cheaper. Nanotechnology will make space travel cheap.

Why use today's materials to design spacecraft? The best spacecraft we can design with today's materials will be pretty lousy and dangerous and expensive. Why bother wasting that money when we could use it to instead research to find better and far cheaper materials?

Ivan said at January 4, 2007 10:11 AM:

Related thoughts on making sure a self-replicating system doesn't go out of control:

How do you ensure people trying to contact a private server remotely are legit?

Public Key Encryption.

If you can engineer a system to be self-replicating, you sure as hell can add the necessary controls so that it only proceeds when an external (probably human) manager says "go".

Think about ICBMs: the technology needed to set off a nuclear reaction on another continent is far more complicated than the technology needed to ensure only the creator can use it.

Ivan said at January 4, 2007 10:13 AM:

By the way, read Phillip K Dick's story "AutoFac".
Automated factories that create consumer goods for a post nuclear exchange society can't be stopped. The appropriate computers to send the stop message have been destroyed.

Great read.

Jeffrey Gordon said at January 4, 2007 12:39 PM:

Just my 2 cents: I think that external reproductive controls only slow things down, but don't ultimately eliminate the problem.

The root problem is error during reproduction (aka mutation). Eventually a mutation occurs (or they accumulate) that eliminates the external constraint. The only way to stop this would be to inspect every replicant to ensure error free duplication has occured. As soon as the poplutaion of replicators becomes large enough, this is simply not feasible.

Therefore, I conclude that external controls or not, if the population is large enough, it will have run away growth. That is, until it encounters the next growth limit (e.g. food supply).

Also, don't count on being able to design a 100% perfect duplication process. The second law of thermodynamics guarantees you won't.

BTW, this occurs all the time in "wet" nanoware. We call it cancer. From a different point of view, it might be called evolution. Depends what your goal is.

Randall Parker said at January 4, 2007 5:43 PM:


The solution is to not store all the instructions needed for replication in the replicating devices. Send out pieces of instruction. Let them do that step. Then send out another piece that overwrites that piece and have them do that next step.

Another idea: Have different kinds of devices do different steps in the replication. Make no one single kind of device that can do all steps. Greater specialization of labor makes the devices easier to control.

Gerald Hibbs said at January 5, 2007 6:13 PM:

The only time I imagined self replicating machines was for space exploration. Have 1000 units take off from the moon. When they reach their destination they make 1000 more that take off to explore -- and so on.

Otherwise, nanobots are going to have, usually, very specialized jobs. Why would you even want them to be self replicating? For personal use, for example, you will have your own nano-cloud that takes care of a host of functions. Everyday you go home, your systems determines bot loss and prints out replacements. They join your cloud and you are ready for the next day.

If we were to go for the ability to maintain your own cloud then would they need to be self replicating? Wouldn't it be wiser to have a percentage of your cloud to be, essentially, builder bots? Again, as losses accumulate the system sends out instructions for the building of replacement nanos on an as needed basis.

There is no question of mutation, you have a central source holding the build instructions. There are still lots of doomsday scenarios possible and you sure as heck don't want to get a virus in there altering your build plans so you are producing something toxic. Hackers of the future will be responsible for lots of deaths.

Engineer-Poet said at January 5, 2007 8:52 PM:

Yes, yes, of course replicators will solve energy problems!  They'll solve a great many current problems (and create new ones).

I have also read "Engines of Creation", and the intervening 20 years have showed me just how long some potentials take to be actualized.  Unfortunately, we need to take care of the next ten years here and we cannot depend on replicators arriving in time.

If success can only be assured by multiple avenues of attack, replicators certainly cannot be our only one.  If a few tens of millions of dollars are enough to push replicator research forward, let's spend them.  But right now we need billions spent on the things we need to feed, clothe and run society until replicators arrive.

James Bowery said at January 6, 2007 10:29 AM:

At the risk of repeating myself:

See here.

Frieta and Drexler both came along with their bogus self-replication approaches at a time when there was a partial self-replication proposal on the table for boostrapping solar power satellites using telepresence requiring no breakthroughs in artificial intelligence. The system design as of 1980 would have a doubling time of 90 days:

O'Neill, Gerard K.; Driggers, G.; and O'Leary, B.: New Routes to Manufacturing in Space. Astronautics and Aeronautics, vol. 18, October 1980, pp. 46-51.

OK, so now Cochrane has caught up somewhat.

However, people are still ignoring the fact that the thing to do _now_ (in addition to using partial self-replication under telepresence management) is invest in The Hutter Prize for Lossless Compression of Human Knowledge.

I know... I was telling people in the early 1980s that Drexler/Frieta wishful thinking was going to dangerously delay the realization of exponentiation of energy resources using more mundane partial self-replication... and the likelihood that people are going to, now, invest heavily in the Hutter Prize is about the same that they will invest in partial self-replication -- even though the Hutter Prize is the way to get complete self-replication.

Cochrane is the new and improved Drexler/Frieta.

Matt S said at January 10, 2007 4:33 PM:

Gee, self-replicating machines that eat rock and collect solar energy. Perhaps they can just put that energy in, say, part of their bodies and we could burn it. Now what would we call these things? Trees? Corn? I don't know, I can't think of anything.

srovnání cen said at July 8, 2008 6:56 AM:

Post nuclear exchange society doesn't sounds great :S

Justin Rion said at April 9, 2015 7:31 AM:

Actually Richard Jones concerning softmachines likes to try to mark his area plus claim it their Drexler proposals were only for 'dry systems.' The dry systems had your advantage inside the 1980s and also early 90's concerning being easier to draw a diagram of to adhere inside a book. They had been also systems that were much more obviously describable. Such as look below is a small seesaw, enable me personally make use of it to explain how one lever then fulcrum method could efforts. Examine that limited scale because of laws and regulations of scaling it could obviously feel displayed in order to posses this particular high level of performance.

Post a comment
Name (not anon or anonymous):
Email Address:
Remember info?

Go Read More Posts On FuturePundit
Site Traffic Info
The contents of this site are copyright ©