Centauri Dreams reader Brian Koester passed along a link to a provocative video last month that spurs thoughts about the nature of interstellar probes. The video is a TED talk delivered by Paul Rothemund in 2007. For those not familiar with it, TED stands for Technology, Entertainment, Design, a conference that began in 1984 and now brings together interesting scientific figures whose challenge is to give the best talk they can on their specialty within the span of eighteen minutes.
I’ve been pondering Rothemund’s talk for some time. You can call this Caltech bioengineer a ‘DNA origamist,’ meaning that he is exploring ways to fold DNA into shapes and patterns. As becomes clear in his presentation, folding DNA into ‘smiley’ faces or maps has a certain wow factor, but once you get past the initial wonder of working at this level, you begin to appreciate how research in DNA nanotechnology points toward self-assembling devices that can be built at the micro-scale.
Molecular Computing to the Stars
And now we’re off to the races, for as Koester noted in his email to me, a small interstellar probe could theoretically create a molecular computer which could then, upon arrival, create electronic equipment of the sort needed for observations. Think of a probe that gets around the payload mass problem by using molecular processes to create cameras and imaging systems not by mechanical nanotech but by inherently biological methods.
A Von Neumann self-replicating probe comes to mind, but we may not have to go to that level in our earliest iterations. The biggest challenge to our interstellar ambitions is propulsion, with the need to push a payload sufficient to conduct a science mission to speeds up to an appreciable percentage of lightspeed. The more we reduce payload size, the more feasible some missions become — Koester was motivated to write by considering ‘Sundiver’ mission strategies coupled with microwave beaming.
The question becomes whether molecular computing can proceed to develop the needed instrumentation largely by tapping resources in the destination system, a process John Von Neumann called ‘interstellar in-situ resource utilization.’ The more in-system resources we can tap (in the destination system, that is), the lighter our initial payload has to be, and yes, that raises countless issues about targeting the mission and the flexibility of the design once arrived to conduct the needed harvesting.
Toward Self-Replication?
What an interesting concept. It’s fascinating to see how far the notion of self-replication has taken us since Robert Freitas produced a self-replicating interstellar probe based on the original Project Daedalus design. That one, called REPRO, would mine the atmosphere of Jupiter for helium-3, just like Daedalus, and would use inertial confinement fusion for propulsion. But REPRO would carry a so-called SEED payload that, upon arrival on the moon of a gas giant, would produce an automated factory that would turn out a new REPRO every five hundred years.
But REPRO would have been massive (each SEED payload would weigh in at close to five hundred tons), with all the challenges that added to the propulsion question. Freitas later turned to nanotech ideas in advocating a probe more or less the size of a sewing needle, with a millimeter-wide body and enough nanotechnology onboard to activate assemblers on the surface of whatever object it happened to find in the destination system.
Now we’re looking at a biological variant of this concept that could, if extended, be turned to self-replication. Rothemund says that he wants to write molecular programs that can build technology. A probe built along these lines could use local materials to create the kind of macro-scale objects needed to form a research station around another star, the kind of equipment we once envisioned boosting all the light years to our target. How much simpler if we can build the needed tools when we arrive?
A Long Leap for DNA Origami
Caltech’s Erik Winfree, who works with Rothemund, gave New Scientist a recent update on where the work stands:
Although the team has so far used the technique to build simple pipes… much more is possible, Winfree says. “Metaphorically, this is similar to how genetic programs within cells direct the growth of an organism.”
Winfree and Rothemund speculate that the technique could provide a way to assemble molecular components into useful structures such as tiny electric circuits. It is also possible to use the self-assembling DNA structures to perform computational tasks, adds Winfree.
“It is very powerful for information processing,” he says. “It’s what’s known as a Universal Turing Machine, which means it can carry out any information processing task.”
Can the method ultimately be extended to serve our interstellar purpose? You can get an overview of Paul Rothemund’s work from the video I linked to above, and also in his paper “Folding DNA to create nanoscale shapes and patterns,” Nature Vol 440 (16 March 2006), pp. 297-302 (available online). Publications from Caltech’s DNA and Natural Algorithms Group are available here.
Finally, on the question of assembly and replication, see Barish et al., “An information-bearing seed for nucleating algorithmic self-assembly,” Proceedings of the National Academy of Sciences Vol. 106, No. 15, pp. 6054-6059 (available online).
Nanomachines as interstellar craft have a lot going for them staring with reduced energy for launch and the large volume with which they could be produced. This latter point means that we don’t have to be very good at aiming them since you can use a shotgun approach with the hope that at least one of them will be on target. Likewise, if even a large percent are destroyed by interstellar dust particles or cosmic rays there’s still the calculable probability that at least some would get through. Also, the very high speeds with which they could be launched means that the travel times would be short.
However, some issues arise. You would probably need lunar manufacturing capability to construct the launch and power equipment needed. This is doable but not in the near term.
For a DNA-based approach would purines and pyramidines be readily available in the exoplanet environment? If not, then the DNA would have to be designed to construct them from CO2, H2O, and N2. This might eventually be doable, but by the time that we achieve that, desktop atom-by-atom production of self-replicating molecules will probably be possible. So we would be facing the possible risk of an ecophagic self-replicating molecule destroying humanity. This means that the DNA interstellar probe should also be able to produce humans, habitat, life-support, etc at destination. But that capability might well be past the time that nanoweapons and accelerating AI are possible.
Finally, how would the DNA probes decelerate? Presumably this would be through a very thin, powered-up superconducting loop. Is that even possible?
One additional point. We use 20 amino acids to produce proteins. There’s probably a reason that we need so many. If we were to try to add functional groups to the DNA origami so that they could, let’s say, hold and break N2 then we would have fundamentally changed the molecular basis of that DNA and we’d need to develop entirely new DNA replication systems which would preserve those functional groups. If the functional groups were to be bound to the nucleosides, it might be pretty difficult because those nucleosides are pointing inwardly towards each other and are in fixed positions. DNA may be good for making gross structures but I think it would be pretty difficult to get them to construct molecules.
Although Rothemund seems to disagree, and strongly, as per his TED talk. I do think the other issue you raised, about decelerating a small probe, is significant. Deceleration issues are often overlooked in probe design, considered as a kind of after-thought, and I’m not sure how we might slow down, say, Freitas’ ‘needle’ probe. Interesting room for speculation there.
How about an organic interstellar probe of any necessary size?
We always think of spacecraft as having to be made of metals and
plastics? Would an organic craft work better in deep space? Could
it have a skin that stops radiation? Perhaps “bladders” of water
or other chemicals to keep the radiation away?
Could we genetically engineer a being/craft that has the mind to
handle such a mission? Maybe artilects will be made of flesh
instead of titanium and do much better in space. It may be much
easier for such a being/craft to reproduce.
Just thinking outside the box.
Hi Paul,
Gregory Matloff recently wrote a paper in JBIS Vol62 2009 called “ElectroMagnetic starship deceleration in a stellar wind”, this is an interesting paper looking at a third option in thrustless interstellar deceleration where he looks into using a starship’s EM field to reflect the destination star’s stellar wind. The other 2 options are EM reflection of interstellar ions and solar sail deceleration (for close rendezvous to the destination star), maybe some of the principles can be used for interstellar nanoprobes…
Cheers, Paul.
I’ve always wondered how these nano-bots, however constituted, would actually go about their job. I like to think of it in terms of larger biological machines that go about this today – like any common animal.
First you need sufficient intelligence and maneuverability in the probe to locate its prey, much as an animal hunts for food. In this case the prey would most likely be some sort of space rock. It may be desirable for the probe to be able to identify the tastiest morsels from a distance so it doesn’t waste time and energy on rocks that are poor in the sort of nutrients it needs. Metals would likely rank high (I am assuming the probe is restricted to physical and chemical manipulation, but not nucleo-synthesis).
Next, the probe must deploy nano onto the rock (think lichen) if it is larger than the probe, or perhaps it can eat the rock whole (think stomach) and have access to a larger range of digestion methods. Digestion is a problem. The nano has to have enough energy on board to complete its task of refining the rock, including deriving energy either from the rock itself or, say, starlight. Digestion is a process of separating out the useful stuff (elements and compounds) from the useless stuff (slag, or waste), and breaking down the useful stuff into simpler molecules that will feed into synthesis.
The raw food then must be transported to synthesis nano that, like a cell, assembles complex molecules and crystals, and which then must be further assembled into the target objects. There may be further intermediate stages, and one of those intermediate stages might be a small factory that can construct the larger units more effectively than nano alone could.
All of this assumes that the mass of the probe is as small as possible and so must acquire the material at the destination for its observation, communications and other equipment. I think we are still quite a ways from being able to build a probe that can do all that and also achieve deceleration as John and Paul have already mentioned. However done, I suspect that a brake will require a substantial increase in probe mass.
There’s also a potential moral issue here: even if the probe is not an all-consuming destructive Von Neumann probe, a device which lands on a planet and starts consuming potentially valuable resources might be looked at by any natives as extreme effrontery on our part or even potentially an act of war. Especially if we and this probe weren’t able to previously identify the presence of those natives due to reasons of scale or extreme biological differences. If the probe started eating living (possibly intelligent) things or real estate simply because they contain the chemicals it needed, that would not merit good neighbor points for us.
Freitas’ REPRO was gargantuan because he maintained a “Daedalus” 0.12c cruising speed, which wasn’t necessary if the reproduction time was ~1,000 years. Doubling the trip-time cut the launch mass back 100-fold. Doubling the trip-time again would cut out a second stage. But that’s just a quibble. Making instruments at destination is a good idea regardless of how it’s actualised.
As for “ecophagy” surely you’ve Freitas’s own paper on the limits of such things happening before being detected and counteracted? My personal thought is that if 4 billion years of an ocean full of evolving biochemistry didn’t make such a super-replicator then human tinkering isn’t going to do it either… BUT I can’t be sure. No one can. The number of possible DNA encoded biomachines is mind-boggling to say the least.
It looks like plants might solve the interstellar communications lag:
http://www.technologyreview.com/blog/arxiv/23581/
So does anyone want to discuss the merits or detractions of a truly
organic, living starship? And I do not mean with a crew aboard.
ljk: Sure, once we figure out if (a big “if”) entanglement is something that can actually be exploited for communication. As to a living starship – do you want it to just be alive or also intelligent?
Both of course, Ron. And yes, I am serious.
A metallic starshiip would need an AI anyway, so perhaps an organic
brain that is conscious and intelligent will be easier to produce.
I was also being serious. An intelligent ship would most likely have problems similar to humans, including boredom, rebellion, depression, madness and so on. It may be possible to somewhat design around that, such as letting it sleep for the majority of the journey, but any real intelligence would ultimately make its own decisions based on its own priorities when it far out of reach. Any intelligence is vulnerable to this, no matter how dedicated to the mission at the outset.
If simply “alive” but not intelligent, a lot can still be accomplished though at a cost of flexibility, especially reacting to unforeseen circumstances.
I am doubtful that there’s a middle ground where you can achieve the benefits of intelligence but without the threat to the original objectives. I don’t expect that whether it’s metallic or organic will matter if it’s truly AI.
Ron S. said:
“I was also being serious. An intelligent ship would most likely have problems similar to humans, including boredom, rebellion, depression, madness and so on. It may be possible to somewhat design around that, such as letting it sleep for the majority of the journey, but any real intelligence would ultimately make its own decisions based on its own priorities when it far out of reach. Any intelligence is vulnerable to this, no matter how dedicated to the mission at the outset.”
I bought up these exact issues back in the thread on Icarus, the successor
to Daedalus. The AI issue is one I think to be as important as the starship
propulsion because the “dumb” computers we have now will not cut it on a
decades or centuries-long mission to star systems light years away from any
help by Earth-based humans or resources that can be accessed along the
way other than the occasional dust grain or hydrogen atom.
Can we develop an AI system that “acts” in such a way that independent
thought is created without actual awareness so that we don’t end up with a
starship that is a basket case while still being able to conduct an interstellar
mission successfully? On the other hand, we may be insulting a superior
mind to think that it would not have already considered all the options
and issues with such a long and potentially hazardous journey and either
opted out or planned measures to cope with its voyage.
Otherwise we will need an “aware” thinking system for all our deep space
missions and whether the AI (or Artilect as the term I prefer) will even
want to leave for another star system on what may be essentially a one-way
suicide mission for it (him/her?) is not a trivial matter. Our first interstellar
space probes will very probably not be coming back to the Sol system, and
even if they could, would a being with superior thought capacities want to
stick around one planetary system with a bunch of talking monkeys with
car keys when it has an entire galaxy it can explore?
The Artilect will have to be made aware before the mission departs (there will
be no epiphanies halfway to Alpha Centauri), so whether it wants to go or not
will be do or die for the mission. As I asked above, can we make a “smart”
AI system that behaves like an independent being without being actually
aware and bringing all that baggage along? And if this Artilect becomes a
being, does it have rights and is our legal system ready to deal with beings
who will be made to essentially serve humanity?
This is my concern that the Icarus team and other engineers thinking
about starships will be focusing so much on the body of the vessel that
they don’t give the same amount of attention to its mind, which will be
no less critical than our minds. Certainly when Daedalus was being planned
in the 1970s, the BIS team didn’t seem to give any consideration to what
they called the “semi-intelligent” computer brain of the star probe after it
zipped through the Barnard’s Star system other than to say, in essence,
“buh-bye”, leaving an aware mind to die a long and slow death in the dark
and cold of interstellar space with nowhere to go as its power systems eventually
shut down.
I also know the BIS team didn’t think at all about what kind of information
package to place aboard Daedalus to tell future finders of the probe who
made it and why it is drifting through the galaxy, but that is another matter
for now.
This brings us back to sending out a starship full of humans, but we run into
similar problems and we may need to build a multigenerational colony ship
to be a success, but that also leads into a host of potential problems.
At this point it seems that the only intelligent species which will have
successful interstellar capabilities are the ones who either live very long
times and don’t mind waiting to get somewhere or they have developed
faster-than-light (FTL) starcraft or know where a useful cosmic wormhole
is located.
For those who think Artilects are still pipe dreams, check out the ideas and
work done by Hugo de Garis so far here:
http://iss.whu.edu.cn/degaris/
And this recent interview with de Garis here:
http://machineslikeus.com/interview-hugo-de-garis.html
I recently read this interview with Nobel Laureate Gerald Edelman, who
has been giving a lot of thought to the idea of consciousness:
http://discovermagazine.com/2009/feb/16-what-makes-you-uniquely-you
I reproduce this relevant quote from page 3 here:
“Eugene Izhikevitch [a mathematician at the Neurosciences Institute] and I have made a model with a million simulated neurons and almost half a billion synapses, all connected through neuronal anatomy equivalent to that of a cat brain. What we find, to our delight, is that it has intrinsic activity. Up until now our BBDs had activity only when they confronted the world, when they saw input signals. In between signals, they went dark. But this damn thing now fires on its own continually. The second thing is, it has beta waves and gamma waves just like the regular cortex—what you would see if you did an electroencephalogram. Third of all, it has a rest state. That is, when you don’t stimulate it, the whole population of neurons stray back and forth, as has been described by scientists in human beings who aren’t thinking of anything.”
If this work is headed in the right direction to real AI and the world of the
Artilect, it will be very very interesting to see what this “mind” does as it
acquires more neurons.
It’s been done already. We’re it!
Okay, lot of ground to cover here.
First off, DNA isn’t really the best choice. PNA (amino acid based nucleic acid) or even GNA (lipid based NA) are stabler and have more raw material around. Granted, a few weeks ago chemists finally solved a natural pathway towards (2 of 4) RNA molecules, but RNA is fairly unstable.
Second, even with stable NAs, biological molecules spontaneously revert in chirality over time, so the probe will need to have an active metabolism and so a hefty energy supply. Cryogenic techniques to avoid this must solve serious biological damage problems.
Third, even with a metabolism, NA based systems are prone to errors (mutations, but also read errors). So the design need to be biologically robust rather than technologically perfect.
And now some notes and nitpicks:
“It is very powerful for information processing,” he says. “It’s what’s known as a Universal Turing Machine, which means it can carry out any information processing task.”
Interestingly, AFAIU DNA _as used in biological systems_ aren’t Turing. It seems that it isn’t necessary to be able to create adaptive systems (genomes of populations learning from their environment, or from them sentient organisms doing the same).
That is of course not the way to go, since (some) Turing systems are easy to understand. (Some, as in C as opposed to, say, Windows. :-)
“We use 20 amino acids to produce proteins.”
No, we don’t. Most cells have AFAIU 22 AAs, two of which are made by post-translational modifications by way of exception codes. There is a lot of contingent tweaking in evolutionary processes, because that is what it is about. Even ‘the’ genetic code isn’t universal, especially mitochondria that have undergone heavily accelerated evolution has a lot of different alternates.
“DNA may be good for making gross structures but I think it would be pretty difficult to get them to construct molecules.”
That’s why RNA come first, and is still used in for example ribosomes to make proteins. Ribosomes, at least without all the bells and whistles of evolution add-ons, can be fairly simple constructs to have incorporated in the probe factory.
Some more odds and ends:
“a device which lands on a planet and starts consuming potentially valuable resources”
Luckily organics are plenty on some asteroids and all comets. The reproductive period would be best spent there, without having to pay the energy and mass motherhood penalty by venturing into yet another massive gravity well soon after braking. Surveying can be done remotely or by sacrificial probes.
“Can we develop an AI system”
I’m fairly certain we can, but it will be fearsomely difficult. It seems that natural minds have been constituted by embodiment, so first one has to provide a hardware-software realization (or at least it’s complicated virtual world equivalent), then emulate the hardware in software. (The later would be great if the AI itself did, of course.)
The reason to believe this and why one can be cautiously optimistic is because recently a biologically faithful model of the cortex have been shown to spontaneously organize to exhibit symbolic thinking:
“After this training, the prefrontal layer had developed peculiar sensitivities to the output. In particular, it had developed abstract representations of feature dimensions, such that each unit in the PFC seemed to code for an entire set of stimulus dimensions, such as “shape,” or “color.” This is the first time (to my knowledge) that such abstract, symbol-like representations have been observed to self-organize within a neural network.
Furthermore, this network also showed powerful generalization ability. If the network was provided with novel stimuli after training – i.e., stimuli that had particular conjunctions of features that had not been part of the training set – it could nonetheless deal with them correctly. This demonstrates clearly that the network had learned sufficiently abstract rule-like things about the tasks to behave appropriately in a novel situation. Further explorations involving parts of the total network confirmed that the “whole enchilada” was necessary for this performance; without an adaptive gating unit, or connecting an additional context layer to the PFC layer (making it equivalent to a simple recurrent network, or SRN) did not demonstrate equivalent generalization.”
So symbolic thinking, i.e. learning without the usual problem of over-training in neural nets, is spontaneously achieved but depends on the overall system and its hardware.
“It looks like plants might solve the interstellar communications lag”
Somehow I don’t think that 5 ps travel time allows for interstellar trips. :-) It doesn’t even allow for neuronal communication tricks, as Tegmark once showed.
Really, this entanglement is an interesting but non-functional consequence of potentially useful quantum coherence. Potentially, since I believe what this untested hypothesis is based on is itself yet not tested. It may be that photosynthesis is all classical because life never stumbled on the few everyday quantum effects one can experience.
Peter Watts’s novel Blindsight included an alien intelligence which was not ‘conscious’ in the sense we understand it. A fascinating story with a genuinely alien alien. And some good speculation too.
Ron said “An intelligent ship would most likely have problems similar to humans, including boredom, rebellion, depression, madness and so on. It may be possible to somewhat design around that, such as letting it sleep for the majority of the journey, but any real intelligence would ultimately make its own decisions based on its own priorities when it far out of reach. Any intelligence is vulnerable to this, no matter how dedicated to the mission at the outset.”
I’m not so sure about AI susceptibility to those ailments. I think we suffer these problems when we immerse our brains into environments which are radically different from the world with which they were evolved to cope. An AI would presumably be ‘evolved’ in exactly the environment that it was designed to work in. We would provide it with a virtual reality in which we throw a bunch of scenarios at it to see how it copes, and (in manner of genetic algorithms) select for ‘breeding’ those AIs that respond well to the appropriate environments. So we’d breed those AIs that cope with (what we would call) boredom, etc.
I take your point that the AI will ultimately develop its own motivations (and of course they may not match up with our desires at all!), but my quibble is with the idea that the psychological problems are “most likely”. I think they’re peculiar to our meat brains and our evolutionary heritage. We should be able to avoid that in our AI descendants.
BTW If you’re interested, Peter Watts’s novel Blindsight is available to read online: http://www.rifters.com/real/Blindsight.htm
Going back to the original concept – why would there have to be any payload at all? Isn’t it conceivable that by the time we have in-situ self assembling probes we could also induce the process remotely with encoded laser/maser energy? The resulting basic molecular templates (à la RNA) could be combined to produce the simplest toolset to bootstrap the rest of the assembly.
Looking at Sirius reflected in a glass patio table a couple of nights ago got me thinking about the idea of extremely long range manipulation of material by light – not with photolithography of course but some sort of manipulation based on polarization, phase modulation, scanning, etc. to create the basic structures. The process would obviously need to be highly adaptive to respond to the target’s material composition and likely require evolved structures to achieve operational compliance.
Lots of TBD’s, but I really like the idea of completely removing the mass and just sending the information to grow the instrumentality along with the minimum energy needed to induce the process.
Pat: “I’m not so sure about AI susceptibility to those ailments.”
What is AI? If any intelligence is of sufficient order to be self-directing and creative to the degree needed for an interstellar mission, I believe it will above the threshold which would make it susceptible. I find it unlikely that there is any ‘middle ground’ between non-self aware, manageable AI which are immune to these problems (or, attributes) and those that are not. I like to think of it in terms of formal logic (which is most probably fundamental to mind), where if a system is strong enough to contain arithmetic it is also provably Godel-incomplete. There is a bright line between the two domains.
I think the idea we can program in certain behaviors harkens back to Asimov’s 3 laws of robotics. I think this is a fiction.