It’s been a week with an exoplanet focus, what with the interesting Kepler results yesterday and the five, or perhaps seven, planets found around the same star by the HARPS instrument. But I can’t close the week without recourse to Seth Shostak’s recent comments on biological versus machine intelligence. Paul Davies took much the same tack in his recent book The Eerie Silence (Houghton Mifflin Harcourt, 2010), arguing that any civilization we encounter will likely be composed of intelligent machines. Shostak thinks SETI should take that seriously.
Searching for Doppelgängers
Right now we’re searching for what Shostak calls ‘doppelgängers of humans’ — i.e., SETI has focused on places that could support life forms that do more or less what we do, which includes not only using radio to communicate, but much broader traits like living for finite lifetimes, following basic biochemical dictates and being subject to evolution. That biases the search toward places that could sustain life as we know it, a reasonable bias in my opinion, but one that may not take our own development into consideration. After all, we may be living in the short window between developing radio and building true artificial intelligence.
Suppose, Shostak asks in this BBC story (with accompanying audio interview), we develop true AI by the end of this century. What would happen next? This is where things get interesting. Shostak:
At some point they may just pick up and leave, at least some of them, maybe most of them… If you’re a machine, you’re interested in only two things I can think of. And that is matter and energy, because those facilitate whatever it is you’re doing. And matter and energy are not in particularly great supply here.
The result: AI lifeforms go to places more suited for their kind of existence, which could include the galactic core or, perhaps, the neighborhood of a hot, young star. Shostak is arguing that we should allocate a small percentage of our observing time — perhaps up to ten percent — for searching in places AI is more likely to call home. Thus far we’ve searched fewer than a thousand star systems intensively, and our all-sky search is of necessity unable to linger on a target. We’re new at the game, in other words, but let’s tune up our target list.
Is Biology the Issue?
The problem with SETI is that we’re forced to make assumptions about how aliens would operate, an issue that continues to bedevil the field today. Recently we’ve looked at the Benford brothers’ call for a different kind of search, one homing in on the kind of interstellar beacons an alien culture would be likely to create. The discipline is rife with new ideas as we try to figure out the basic parameters that any intelligent species would have to possess in our galaxy. But getting into an alien mind, much less an artificial one, is tricky business. The best we can do is build on our knowledge of physics and extrapolate a line of rational behavior.
The Benfords extrapolate from both physics and economics to argue that an interstellar beacon will likely use short, powerful bursts rather than continuous broadcasts. But SETI has lacked the ‘long stare’ needed to find such a signal. To me, the issue is less AI vs. biology than it is continuous survey vs. pinpoint search. The SETI League’s Project Argus, discussed in these pages recently, is an attempt to set up 5000 amateur receiving stations to implement the ‘long stare.’ It would be a low sensitivity survey, but as the cost of equipment drops and its power increases, it should become possible to implement at the amateur level, and it could be a powerful adjunct to more sophisticated (and focused) searches.
Methods like these could detect an alien beacon, whether built by machine or biological beings, out to several hundred light years, with the sphere of detection growing as we replace older stations with newer technology. They’re a great complement to higher-powered instruments. If we’re looking for beacons, a continuous, high-sensitivity stare along the galactic plane is a sensible way to proceed. But there’s a place for minimal assumptions and broad coverage too, and the advantage of an all-sky survey is that it takes what it finds, which might involve the kind of surprises SETI is made for.
Here’s a thought: Is it possible that Kepler may inadvertently find a “Benford Beacon”?
Ahh that old horse, intelligent machines, is dragged out and dusted off again. Yes, machines with subtle and extremely flexible operating programs are entirely possible to build but at the end of the day they are simply machines, not individually unique self aware thinking beings. We neither have now or can even envisage the technology to produce pure machine intelligence. I will grant the possibility of an organic/non-organic hybrid but that of course falls under the realm of the first statement.
It’s possible that machine intelligence would appear as an unintended side effect of some other developments. In fact any machine exhibiting it might immediately be shut down because its behavior was so unpredictable.
Interesting take on SETI and machine intelligence. Well, if they do exist, we may need to adopt a contrarian “avoid the water” search strategy.
I can’t help but think the last place an intelligent amalgam of metal and circuitry would want to be is on a humid, stormy, corrosive, biology-infested world — kind of like Earth!
Avoid The Water! Thanks Scott. You’ve finally put the two must-bes together for me: that the galaxy has long since been heavily populated, and that no aliens have shown up here. They’re simply not interested in Earth or us. Well then, what of the rest of the solar system? Maybe the asteroids? Or Mercury?
@JDE
Well, if you accept that a human mind results from a brain, which is a wet electro-chemical organ, and not some mysterious soul or spirit, I think AI is almost inevitable. I think we understand the basic chemistry and physics of the brain well, and we can simulate portions of brains (IBM research, etc.). In principle you can simulate a whole brain on a powerful computer. If you can do that, then you should be able to simulate a mind.
I’m not saying it is easy, but I see no real problem with AI being feasible in the next 1,000-10,000 years (in other words, shortly after the invention of radio). Even my phone has voice recognition, and that type of thing was an “AI” research project not more than 15 years ago.
I really think Davies and Shostak have sound reasoning. Matter and energy are the ultimate currencies, so these would be interesting issues in focusing our search. But energetic things already attract lots of attention in astronomy. What would it take to data mine observations (I’m sure we’ve got some good data gathered over years) of the galactic core for SETI research?
JDE: By definition, an AI is not “simply a machine”, but an “individually unique self aware thinking being”, one which happens to be built from metal and silicon rather than water and carbohydrates. You may claim this is impossible, but I don’t think you can back that up with rational arguments or evidence.
NS: We work every day with human beings, to great advantage, and without shutting them down for unpredictability (well, rarely, anyway). I don’t think machines would be treated any different.
Scott G: Remember that Earth is made mostly of Metal and ceramics, the perfect raw materials for building machines. The thin layer of pesky stuff can easily be peeled off…
Eric:
I do not think that brain simulation is the right approach to AI. It tells you much more about the brain than about the mind, and remember that the brain is the part AI wants to do away with. Simulating the brain to understand the mind is like trying to understand a computer program by simulating the processor that is running it.
AI sits somewhere in the middle between psychology, neuroscience, and computer science and has to develop its own paradigms. I think psychology is actually the most relevant discipline here, as the science of the mind. You could see AI as the engineering discipline arising out of psychology.
That said, in my opinion progress has been phenomenal so far and I expect personality simulations that are nearly indistinguishable from a human at a keyboard to become feasible before the century is out. It will be a wonderful development, because finally you will be able to argue with and yell at customer support again….
My take on it is that the more eyes and ears we have on the sky, the more we’ll find. Random sweeps are not sufficient for finding whatever may be out there. The closer we get to continuous observation throughout the sky, the better it will be for SETI as well as other avenues of research.
It is important that we not get caught in anthropomorphism – while aliens will be bound within the same physical laws, there is no reason to assume they will approach communication technology in the same way as us.
I got carried away with Scott’s idea of “avoiding the water” to find evidence of a galactic machine civilization. So I checked to see if we had found anything odd about Mercury (plenty of energy and material for machines). What I discovered would make the perfect beginning for a science fiction novel: In 2004 the US sent the MESSENGER spacecraft to Mercury. Last year, in September, it made its third and closest flyby before it settles into an orbit next year. It carried cameras that could resolve objects of 59 feet, so I’m sure there was some excitement in mission control. But “sometime during the closest approach the spacecraft entered safe mode. Although this had no effect on the trajectory necessary for later orbit insertion it may have resulted in the loss of science data and images that were planned for the outbound leg of the fly-by. The spacecraft had fully recovered by about 7 hours later.” Too much! This from Wikipedia. Well, if they are there, I suppose anything they would do would seem like magic to us, including manipulating our craft. The other very odd thing that was discovered during this mission is “large amounts of water present in Mercury’s exosphere. “Nobody expected that. I don’t know a single person that did. We were astonished, just astonished,” Zurbuchen stated.” I’m not sure that supports my idea, but I thought I’d throw it in as a matter of interest.
I wrote a series of novels called the Galactic Center series, that assumes most intelligence in the galaxy is machine. There’s a whole timeline worked out, starting with our near future, starting with In the Ocean of Night.
So machine SETI is a natural outcome of such thinking.
If AI civilizations have a truly long scale perspective, they might not do SETI at all, but rather explore with self-reproducing machines, being very careful to disguise their presence.
Still, such ideas are hard to check. That’s why we worked on the Benford beacon approach, constrained by economics and physics.
Gee, matter and energy…sorta narrows it down eh?
better to think of intelligent machines as hybrids. it is likely that biology will dovetail nicely with machines. couple that assumption with projections in physics and you may well glimpse future potentials for “life”.
What is the best way to worm through a worm hole? Hint: think like a little tiny wormy thing.
Interesting assumption that machine intelligences would move to where matter and energy and abundant, as if replication in machines was teh same as living systems.
I would argue that machine intelligences might go to where new information is abundant, exactly those messy, biologically infested planets. Which means that we should be looking inside the solar system, possibly even much closer to home.
The problem with SETI is that it makes assumptions based on our current cultural ideas and level of technology. Each disappointment just ups the ante to the current level of technology and starts again. Thus we abandon the idea that long radio broadcasts are needed for primitive civilizations to detect and search for the next minimum technology level. We are now at ultra short optical laser bursts, and no doubt that will be obsoleted in 10 – 20 years in favor of the next technology the aliens might use…
Hey, I’m interested in energy and materials!
Hmm well Eniac I pose the same challenge to you. Back up your claims that machine intelligence will appear. Modeling sections of the brain with machine code is the same thing as modeling the human body with clay. Yep you can get purdy results but at the end of the day you still only have a lump of clay.
This whole discussion could easily evolve into a metaphysical angst festival with definitions of “what is life” being thrown around.
JDE: Fair enough. Despite all the progress that has been made, there is currently no clear path towards creating AI, or even a consensus on what exactly we mean by that. In the absence of evidence or rational argument for either, I would rather assume something is possible than impossible, but that is just my optimistic nature and can-do attitude…
It seems to me that “machine” AI is a thing of the past now. Manipulated DNA is the “AI” of the future, with easy, unlimited sexual reproduction.