Robots in Space
In this posting I will continue with the critique of manned exploration. In the next posting I will include a response that aims to make a case for manned exploration. Before getting to the case for robots in space, I should mention, in connection with the previous posting, that the newest think about teleoperators will have the operator wear a sort of exoskeleton. As the exoskeleton moves, so does the machine in outer space.
The other hope of the opponents of manned spaceflight is the development of intelligent machines. As those critics constantly remind us, computers can already perform better than humans in several areas. One such area is in geology itself – apparently in contradiction to my earlier remark – where expert programs do a better job in the exploration of underground oil deposits. All we need to do is develop expert programs about the solar system and we will produce robot spacecraft that can do the exploration for us. It may be more heroic to do it with humans, but, according to the critics, at those prices we can afford a bit less heroism and a lot more common sense. Some NASA officials fear that unmanned programs will not enjoy as much public support. In their view, the public derives great satisfaction from the vicarious participation in the grand adventure of space exploration. But opponents of manned exploration argue that robots will permit vicarious participation by the general public in the unraveling of the mysteries of the solar system. We have already experienced that pleasure with the Viking landings and the recent rovers on Mars, with the Voyager missions to Jupiter and Saturn, the Galileo spacecraft, and more recently with Cassini’s visit to Saturn and Titan.
There is no question that our exploration of the solar system will be more fruitful the more independent and flexible our spacecraft become. But we should not turn this reasonable wish into wishful thinking. The truly successful expert programs are very few; and the successful ones have very limited applications. The spacecraft cannot take expert programs for every contingency unless the mission is rather simple and we have a fair idea of what those few contingencies may be. An expert program, whether for diagnosing diseases or finding oil, works by compiling a large set of techniques and rules of thumb used by the human experts in the field. The programmer discovers what rankings and relative values those human experts give to those techniques and rules of thumb in typical cases of application. Such a program can thus do a better job than an individual human expert because the computer can keep track of a larger number of considerations. But when the situation is not typical, when it demands different rankings and values, as it is normally the case with the scientific exploration of the unknown, then the expert program soon exhausts its usefulness. Nor can we program in advance those different assignments of rankings and values, precisely because their appropriateness will be determined by unknown circumstances. This is not to say that human beings seldom make mistakes in the way they grasp the situation. Nonetheless, their flexibility does give them an edge where they meet an open-ended environment.
This problem of flexibility is perhaps the greatest barrier to artificial intelligence. There are many programs that perform very well in restricted domains, but no one has an inkling of how to make a program of general application. All too often what from a distance seems to be a difference in degree that can be overcome with larger computing power and memory storage, up close becomes an insuperable difference in quality. The things that computers cannot do are those like using language or going shopping that come so naturally for even the dullest of human beings. Live intelligence constructs a world for itself, i.e., “interprets” the world as it interacts with it. But being able to tell that much does not amount to knowing how it all works, and thus we are certainly in no position to provide electronic equivalents. We have no idea how to make a computer with the world smarts of a dodo. Newspaper headlines about the wonderful things robots will be able to do in ten years or less are simply pipe dreams that cannot be backed up by any actual research in artificial intelligence.
Whether this barrier can be overcome in principle I will not discuss here. For our purposes, the important consequence is that we have no reason to suppose that it will happen in the next decade or so. It is true that we have often achieved what pundits had declared impossible. The space program is one of our very best examples of that. But the history of our scientific civilization is also full of projects that we later discovered could not be realized. Among those projects we should include the proposals made in their youth by Tsiolkovsky, Goddard, and Oberth, the three pioneers of space flight, for a machine that could lift itself into orbit by its self-generated centrifugal force. As it turned out, the device violated the law of conservation of momentum.
In any event, the success of artificial intelligence is not likely to come soon enough to contribute with robots what humans have to offer now to the progress of space science. Where it is inconvenient for humans to go, we must settle for what robots and teleoperators can do. But where we already know that humans can deliver the goods, it is not reasonable to snub them in favor of an uncertain technology. These remarks are not intended to argue against the development of more sophisticated teleoperators and robotics. On the contrary: There are many environments where humans cannot yet go, and will not go for a long time, and others where they should never go. If machines can go in our stead to those environments, we are so much further ahead.
Nevertheless, opponents of manned flight take a look at the costs of the International Space Station (about $130 billion) and point out that a lot of space science could be done for that amount. For example, the NEAR mission to investigate Eros, an asteroid that comes as close to the Earth as 14 million miles, had a price tag of about $211.5 million dollars, which is pretty standard nowadays. Apart from its scientific value, this mission may someday allow us to figure out how to divert from our planet a similar asteroid. It seems incongruous that for the price of one space station we could fly instead between 400 and 500 interplanetary missions!
Let us consider some of the important missions, scientifically and otherwise, whose funding has been affected by the diversion of monies into what some believe is a manned sinkhole in the sky:
- NASA’s “system of environmental satellites is at risk of collapse” because the agency has shifted to the Shuttle and the Space Station $600 million from the Earth sciences.
- NASA, for similar budgetary reasons, has downsized the next generation of the National Polar-Orbiting Operational Environmental Satellite System. In particular, it has stripped out “instruments crucial to assessing global warming, such as those that measure incoming solar radiation and outgoing infrared radiation.”
- For $100 million of fine-tuning the Large Synoptic Survey Telescope (LSST) we could identify 90 percent of asteroids between 100 and 1000 meters in size. And since the LSST is Earth-bound and thus is limited (can spot the asteroids the come closest only at dusk or dawn, when the sun’s glare may obstruct our vision of them), for $500 million we could place in orbit around the sun an infrared telescope that “could pick up essentially every threat to Earth.”
- For $400 million, the proposed Don Quijote would fire a 400 kg projectile into a small asteroid to see how it affects its trajectory. This would begin to help us figure out how to deflect asteroids bound to collide with our planet.
- An orbiter around Europa could determine once and for all whether that Jovian moon really has an ocean. A wandering hot-air balloon in the atmosphere of Titan can tell us whether there are traces of self-organization by the organic substances found there by the Cassini mission. Given NASA’s need to pump money into manned exploration, the agency will have to choose between these two missions. Shouldn’t we do both?
These examples were current as of 2008; by now new cuts in space budgets will probably make the situation worse. Nevertheless, they still give us an inkling of how far space science could go if not for our mania to send humans into space.
When the space station was first proposed, most space scientists feared, with good reason, that, on the whole, the space station was going to take money away from space science. I say with good reason because that is exactly what happened during the construction of the Space Shuttle. As the new vehicle could not be brought in under budget, the space sciences suffered a double jeopardy. First, their funds for many science projects were transferred to the shuttle. And then many experiments were not performed because they had been scheduled to go in the shuttle but the shuttle was not ready. The two shuttle disasters made the situation a lot worse. Money has and will continue to be drained from the space sciences to keep astronauts flying (to accomplish very little science by comparison). It was quite proper for scientists to want to ensure that a commitment to the space station would not be underwritten on the back of space science. Now we can see that their worst fears have been realized.
According to the bioengineer and NASA adviser Larry Young, “NASA always uses research as justification for its large manned missions, but once they are under way the engineering, political, and fiscal factors take over and the science constituency is often cast aside.” Weinberg is far bitter: Of five missions proposed to challenge and expand Einstein’s General Theory of Relativity, only one is likely to survive. "This is at the same time,” he says, “that NASA's budget is increasing, with the increase being driven by what I see on the part of the president and the administrators of NASA as an infantile fixation on putting people into space, which has little or no scientific value." For Weinberg, what has happened to the Beyond Einstein program reminds him of the time when the most grandiose particle-physics project, the Superconducting Super Collider, which was being built in Texas, was scrapped by Congress because funds were needed to build the International Space Station.
It should not be surprising, then, that a great many space scientists are now opposed to the new proposals to send humans back to the Moon and on to Mars. And we cannot blame them for, as we have seen, money is already beginning to drain from unmanned exploration.
As for the disadvantages of teleoperators and robots, the opponents of manned exploration point out, we can send dozens of dumb robots or clumsy tele-operated contraptions to take on the sundry jobs a human could theoretically do in space. Certainly, we will fail far more often, but the failures will not be as costly or devastating; we will save money; we will get the job done; and we will be forced to improve our science and technology. It is not just that we can try with machines again and again until we get it right, but also that we can divide the tasks humans would have performed into many simpler tasks and then try to accomplish those with swarms of new machines.
And let us not forget that machines have traveled tens of thousands of times further than humans have ever gone. What sense does it make to restrict exploration to dipping our toes when we could swim across the English Channel?
 Hoegland, John. Artificial Intelligence: The Very Idea, MIT Press, 1985.
 This point was made as early as 1972 by Hubert Dreyfus in his What Computers Can’t Do: A Critique of Artificial Reason, Harper & Row.
 This and most of the examples that follow are taken from George Musser, “5 Essential Things to do in Space,” Scientific American, October 2007, pp. 75. The present quote is from p. 70.
 Ibid, p. 71.
 Science, Vol. 310, 25 November 2005, p. 1245.
 SPACE.com, op. cit.