Collected Essays
Page 24
The notion of robot A-Life interests me so much that I’ve written several science fiction novels about it. As will be discussed in a section below, The Hacker and the Ants talks about how one might use a Virtual Reality world in which to evolve robots.
In Software, some robots are sent to the moon where they build factories to make robot parts. They compete with each other for the right to use the parts (natural selection), and then they get together in pairs (sex) to build new robots onto which parts of the parents’ programs are placed (self-reproduction). Soon they rebel against human rule, and begin calling themselves boppers. Some of them travel to Earth to eat some human brains—just to get the information out of the tissues, you understand.
In Wetware, the boppers take up genetic engineering and learn how to code bopper genomes into fertilized human eggs, which can then be force-grown to adult size in less than a month. The humans built the boppers, but now the boppers are building people—or something like people.
At the end of Wetware, the irate humans kill off the boppers by infecting their silicon chips with a biological mold, but in Freeware, the boppers are back, with flexible plastic bodies that don’t use chips anymore. The “freeware” of the title has to do with encrypted personality patterns that some aliens are sending across space in search of bodies to live upon.
In my most recent book of this series, Realware, the humans and boppers obtain a tool for creating new “realware” bodies solely from software descriptions of them.
Real Robots
After such heady science fiction dreams, it’s discouraging to look at today’s actual robots. These machines are still very lacking in adaptability, which is the ability to function well in unfamiliar environments. They can’t walk and/or chew gum at the same time.
The architecture for most experimental robots is something like this: you put a bunch of devices in a wheeled can, wire the devices together, and hope that the behavior of the system can converge on a stable and interesting kind of behavior.
What kind of devices go in the can? Wheels and pincers with exquisitely controllable motors, TV cameras, sonar pingers, microphones, a sound-synthesizer, and some computer microprocessors.
The phenome is the computation and behavior of the whole system—it’s what the robot does. The robot’s genome is its blueprint, with all the interconnections and the switch-settings on the devices in the wheeled garbage can, and if any of those devices happens to be a computer memory chip, then the information on the chips is part of the genome as well.
Traditionally, we have imagined robots as having one central processing unit, just as we have one central brain. But in fact a lot of our information processing is down out in our nerve ganglia, and some contemporary roboticists are interested in giving a separate processor to each of a robot’s devices.
This robot design technique is known as subsumption architecture. Each of an artificial ant’s legs, for instance, might know now to make walking motions on its own, and the legs might communicate with each other in an effort to get into synch. Just such an ant (named Atilla) has been designed by Rodney Brooks of MIT. Brooks wants his robots to be cheap and widely available.
Another interesting robot was designed by Marc Pauline of the art-group known as Survival Research Laboratories. Pauline and his group stage large, dadaist spectacles in which hand-built robots interact with each other. Pauline is working on some new robots which he calls Swarmers. His idea is to have the Swarmers radio-aware of each other’s position, and to chase each other around. The idea is to try to find good settings so as give the Swarmers maximally chaotic behavior.
In practice, developing designs and software for these machines is what is known as an intractable problem. It is very hard to predict how the different components will interact, so one has to actually try out each new configuration to see how it works. And commonly, changes are being made to the hardware and to the software at the same time, so the space of possible solutions is vast.
Telerobotics
For many applications, the user might not need a robot to be fully autonomous. Something like a remotely operated hand that you use to handle dangerous materials is like a robot, in that it is a complicated machine which imitates human motions. But a remote hand does not necessarily need to have much of an internal brain, particularly if all it has to do is to copy the motions of your real hand. A device like a remote robot hand is called a telerobot.
Radioactive waste is sometimes cleaned up using telerobots that have video cameras and two robotic arms. The operator of such a telerobot sees what it sees on a video screen, and moves his or her hands within a mechanical harness that send signals to the hands of the telerobot.
I have a feeling that, in the coming decades, telerobotics is going to be a much more important field than pure robotics. People want amplifications of themselves more than they want servants. A telerobot projects an individual’s power. Telerobots would be useful for exploration, travel, and sheer voyeurism, and could become a sought-after high-end consumer product
But even if telerobots are more commercially important than self-guiding robots, there is still a need for self-guiding robots. Why? Because when you’re using a telerobot, you don’t want to have to watch the machine every second so that the machine doesn’t do something like get run over by a car, nor do you want to worry about the very fine motions of the machine. You want, for instance, to be able to say “walk towards that object” without having to put your legs into a harness and emulate mechanical walking motions—this means that, just like a true robot, the telerobot will have to know how to move around pretty much on its own.
Evolving Robots
I think Artificial Life is very likely to be a good way to evolve better and better robots. In order to make the evolution happen faster, it would be nice to be able to do it as a computer simulation—as opposed to the building of dozens of competing prototype models.
My most novel, The Hacker and the Ants, is based on the idea of evolving robots by testing your designs out in Virtual Reality—in, that is, a highly realistic computer simulation with some of the laws of physics built into it.
You might, for instance, take a CAD model of a house, and try out a wide range of possible robots in this house without having to bear the huge expense of building prototypes. As changing a model would have no hardware expense, it would be feasible to try out many different designs and thus more rapidly converge on an optimal design.
There is an interesting relationship between A-Life, Virtual Reality, robotics, and telerobotics. These four areas fit neatly into Table 3, which is based on two distinctions: firstly, is the device being run by a computer program or by a human mind; and, secondly, is the device a physical machine or a simulated machine?
Mind
Body
Artificial Life
Computer
Simulated
Virtual Reality
Human
Simulated
Robotics
Computer
Physical
Telerobotics
Human
Physical
Four Kinds of Computer Science
Artificial Life deals with creatures whose brains are computer programs, and these creatures have simulated bodies that interact in a computer-simulated world. In Virtual Reality, the world and the bodies are still computer-simulated, but at least some of the creatures in the world are now being directly controlled by human users. In robotics, we deal with real physical machines in the real world that are run by computer programs, while in telerobotics we are looking at real physical machines that are run by human minds. Come to think of it, a human’s ordinary life in his or her body could be thought of as an example of telerobotics: a human mind is running a physical body!
Memes
In the wider context of the history of ideas, one can observe that certain kinds of fads, techniques, or religious beliefs behave in some ways like autonomous creatures which live and reproduc
e. The biologist Richard Dawkins calls these thought-creatures memes.
Self-replicating memes can be brutally simple. Here’s one:
The Laws of Wealth:
Law I: Begin giving 10% of your income to the person who teaches you the Laws of Wealth.
Law II: Teach the Laws of Wealth to ten people!
The Laws of Wealth meme is the classic Ponzi pyramid scheme. Here’s another self-replicating idea system:
System X:
Law I: Anyone who does not believe System X will burn in hell;
Law II: It is your duty to save others from suffering.
Of System X, Douglas Hofstadter remarks, “Without being impious, one may suggest that this mechanism has played some small role in the spread of Christianity.”
Most thought memes use a much less direct method of self-reproduction. Being host to a meme-complex such as, say, the use of language can confer such wide survival advantages that those infected with the meme flourish. There are many such memes with obvious survival value: the tricks of farming, the craft of pottery, the arcana of mathematics—all are beneficial mind-viruses that live in human information space.
Memes which confer no obvious survival value are more puzzling. Things like tunes and fashions hop from one mind to another with bewildering speed. Staying up to date with current ideas is a higher-order meme which probably does have some survival value. Knowing about A-Life, for instance, is very likely to increase your employability as well as your sexual attractiveness!
* * *
Note on “Life and Artificial Life”
Written 1992.
Appeared in the Artificial life Lab manual, Waite Group Press, 1993.
There are a number of very comprehensive anthologies of technical and semi-technical papers that have been presented at conferences on Artificial Life. The first conference was held at Los Alamos, New Mexico, in 1987, and its papers appear in C. Langton, ed., Artificial Life, (Addison-Wesley, 1989). A good popular book on A-Life is: Steven Levy, Artificial Life: The Quest for a New Creation, (Pantheon Books, 1992).
I was employed as a “Mathenaut” in the Advanced Technical Division at Autodesk, Inc., from the August, 1988 to September, 1992. While I was there, I worked on CA Lab, on James Gleick’s CHAOS: The Software, on the Autodesk Cyberspace Developer’s Kit, and on a solo project with the working title Boppers. In 1992 Autodesk’s stock went down, and, as I mentioned earlier, they laid off many of the people in the Advanced Technical Division—including me. But they let me keep the rights to my Boppers code, and I got it published as a package called Artificial Life Lab. It’s out of print now, but the Boppers program, the Boppers source code and the complete Artificial Life Lab manual are available on my website.
I really enjoyed my time at Autodesk, but I wasn’t doing much writing while I was there. It was good to come back to the slower pace of academic life. By the end of my four years in the software industry pressure-cooker I felt a like an undercover agent who has forgotten his real identity and has started to believe his cover story. Regarding my return, I had a mental image of a jeep whining up a hill along a wire fence at some Iron Curtain border. The jeep stops, two men raise up a tightly wrapped canvas sack and throw it over the fence, the jeep speeds off. The long canvas bag twitches, unfolds, and there I am, back in the land of literature.
A Note On Synthetic Biology
The SynBio approach is onto something big—a new version of nanotechnology, which is the craft of manufacturing things at the molecular scale. SynBio’s plan is to capitalize on the fact that biology is already doing molecular fabrication all the time. What might happen if we repurpose biology to our own ends?
One big worry is what nanotechnologists call the “gray-goo problem.” What’s to stop a particularly virulent SynBio organism from eating everything on earth? My guess is that this could never happen. Every existing plant, animal, fungus and protozoan already aspires to world domination. There’s nothing more ruthless than viruses and bacteria—and they’ve been practicing for a very long time.
The fact that the SynBio organisms are likely to have simplified Tinkertoy DNA doesn’t necessarily mean they’re going to be faster and better. It’s more likely that they’ll be dumber and less adaptable. I have a mental image of germ-size MIT nerds putting on gangsta clothes and venturing into alleys to try some rough stuff. And then they meet up with the homies who’ve been keeping it real for a billion years or so.
Now let’s look at the upside. Donning the funhouse spectacles of science fiction, I envision a wide range of biotech goodies.
Every child is likely to want a pet dinosaur, and this will be easily managed once the online Phido Pet Construction Kit is up and running. Of course, if you prefer something cuddly, you can design a special dog with red polka dots.
Rather than mining for ore, why not let plants use their roots to extract minerals from the ground? Sow a handful of Knife Plant grain over a dumpsite, and before long you’ll have what looks like corn—but with a cob-handled steel knife in each ear.
Why bother building houses when you can get a Giga Gourd seed? The seed is the size of a pizza and grows very fast. Push it into wet, fertile ground and stand back. In a few days you’ll have a big, hollow home with plumbing and wiring grown right into the walls, which come complete with transparent window patches.
Of course, people will want to start tweaking their own bodies. Initially we’ll go for enhanced health, strength and mental stability, perhaps accelerating the pace of evolution in a benign way.
But, feckless creatures that we are, we may cast caution to the winds. Why would starlets settle for breast implants when they can grow supplementary mammaries? Hipsters will install living tattoo colonies of algae under their skin. Punk rockers can get a shocking dog-collar effect by grafting on a spiky necklace of extra fingers with colored nails. Or what about giving one of your fingers a treelike architecture? Work ten two-way branchings into each tapering fingerlet of this special finger, and you’ll have a thousand or so fingertips, with the fine touch of a sea anemone.
It’s easy to imagine grafting an electric eel’s electromagnetic sensitivity into our brains so we can pick up wireless signals. There’d have to be an off switch, of course, but the net effect could be amazing. We’d have true telepathy, and the ability to form group minds.
As the technology of brain-to-brain contact improved, you’d no longer need to send someone every detail of a plan, a memory or a design. Instead you could send something like a mental Web link, allowing those you invite to simply view your thoughts right in your own mind.
The biggest problem with manned spaceflight is the immense mass of the requisite life-support systems and radiation shielding. What if the truly determined astronauts could transform themselves into tough, spindle-shaped pods that could sail endlessly through empty space, nourishing themselves with solar radiation and directing their journey with the exhalations of their ion jets?
One last thought. Suppose it were possible to encode a person’s memory and personality into a single, very large, DNA-like molecule. Now suppose that someone turns himself into a viral disease that other people can catch. If I were you—sneeze—oh, wait, I guess I am. Are we completely agreed?
* * *
Note on “A Note on Synthetic Biology”
Written in 2007.
Part of a Newsweek article on “Synbio,” May 23, 2007.
This was an odd little assignment where a reporter phoned me up and offered me a nice sum of money for writing a very short article. It’s pretty easy for me to write these kinds of articles, as I have so much material that I can draw snippets from.
Mathematica: A New Golden Age of Calculation
Back in elementary school, we learned procedures, or algorithms, for doing arithmetic with pencil and paper. (Remember “borrowing”?) As adults, we tend to not use our painfully wetware-programmed arithmetic algorithms because most of us have ready access to machines that can do the algorithms by the
mselves. You might occasionally add two or three numbers, but if you have some multiplying or dividing to do, you’re going to search your desk or your desktop for a calculator.
Mathematics doesn’t stop at arithmetic. If you moved further on in school mathematics, you learned more and more algorithms; things like plotting the graph of a straight line, factoring a quadratic equation, and multiplying matrices; maybe you even got to calculus and learned about differentiation and integration. As adults, most of us never need to solve these kinds of problems at all, but if you did have to solve them on a regular basis, what would you do? Chances are you’d get hold of a computer running some kind of computer algebra program.
The oldest such package, called Macsyma, was born at MIT in the 1970s. An original impetus for the project was to help physicists work with formulae that were simply too long and complicated for the human mind—things like the hundred thousand algebraic terms in (you should pardon the expression) the Ricci tensor used in the spacetime field equations of Einstein’s General Theory of Relativity. By the 1980s, Macsyma was like a potbound plant, limited by its design’s restriction to the use of only one megabyte of RAM. Though Macsyma was eventually rewritten, other new computer algebra systems arose to take most of its market. The new programs included Maple (also sold as MathCAD) and—the most expensive and ambitious of them all—Mathematica.
How exactly does one use Mathematica? The shrink-wrap contains a seriously fat user’s guide by Wolfram and a CD with a powerful graphically-interfaced program that runs on virtually every computer platform. You type in any mathematical expression you like and, depending on what you ask for, Mathematica might respond with an algebraically simplified version of the expression, a calculation of expression’s numerical value, a huge data-base table of numbers, or a graph illustrating the expression’s range of values. The graphs can be colored and three-dimensional. With Mathematica and an hour’s practice, a college student can solve any and all the problems in a standard algebra or calculus book.