Climbing Mount Improbable
Page 26
But where does all this complaisant copying and instruction-executing machinery come from? It doesn't just happen. It has to be made. In the case of computer viruses, the machinery is made by humans. In the case of DNA viruses, the machinery is the cells of other creatures. And who manufactures those other creatures, those humans and elephants and hippos whose cells make life so easy for viruses? The answer is, other self-copying DNA manufactures them. The DNA that ‘belongs’ to the humans and the elephants. So, what are big creatures like elephants and cherry trees and mice? (I say ‘big’ because even a mouse, from a virus's point of view, is very very big.) And for whose benefit are mice and elephants and flowers put into the world?
We are closing in on a definitive answer to all questions of this kind. Flowers and elephants are ‘for’ the same thing as everything else in the living kingdoms, for spreading Duplicate Me programs written in DNA language. Flowers are for spreading copies of instructions for making more flowers. Elephants are for spreading copies of instructions for making more elephants. Birds are for spreading copies of instructions for making more birds. The cells of an elephant cannot tell whether the instructions they are slavishly obeying are virus instructions or elephant instructions. As in the case of Tennyson's Light Brigade when someone had blundered, ‘Theirs not to make reply, their's not to reason why, their's but to do and die’.
You will understand that I am using ‘elephant’ to stand for all large, autonomous creatures, for flowers or bees, for humans or cactuses, for bacteria even. The virus instructions, as we have seen, are saying ‘Duplicate me’. What are the elephant instructions saying? This is the main insight that I wish to leave you with at the end of the chapter. Elephant instructions are also saying ‘Duplicate me’, but they are saying it in a much more roundabout way. The DNA of an elephant constitutes a gigantic program, analogous to a computer program. Like the virus DNA it is fundamentally a Duplicate Me program but it contains an almost fantastically large digression as an {272} essential part of the efficient execution of its fundamental message. That digression is an elephant. The program says: ‘Duplicate me by the roundabout route of building an elephant first.’ The elephant feeds so as to grow; it grows so as to become adult; it becomes adult so as to mate and reproduce new elephants; it reproduces new elephants to propagate new copies of the original program instructions.
You can say the same about bits of creatures, too. The peacocks beak, by picking up food that keeps the peacock alive, is a tool for indirectly spreading instructions for making peacock beaks. The male peacock's fan is a tool for spreading instructions for making more peacocks’ fans. It works by being attractive to peahens. It is good at picking up peahens while the beak is good at picking up food. Males with the most beautiful fans will have the most children to pass on copies of fan-beautifying genes. That is why peacock fans are so pretty. The fact that they are pretty to us is an incidental by-product. The peacock's fan is a gene spreader and it works via peahens’ eyes.
Wings are tools for spreading genetic instructions for making wings. In the peacock's case they make their mark as gene preservers especially when the bird is surprised by a predator and shoots briefly into the air. Plants manage something akin to flight organs for their seeds (Figure 8.6), but in spite of this most people would probably not be happy to use the word ‘flying’, in its true sense, for plants. Plants, it seems, don't fly, and they don't have wings.
But wait! From a plant's point of view, it doesn't need wings of its own if it has bees’ wings, or butterflies’ wings, to do the job for it. In fact, I wouldn't mind calling the wings of a bee plant wings. They are organs of flight that are used, by the plant, to ferry its pollen from one flower to another. Flowers are tools for getting plant DNA into the next generation. They work like peacocks’ fans, but instead of attracting peahens they attract bees. Otherwise there is no difference. Just as a peacock's fan works, indirectly, on the leg muscles of the peahen, causing her to walk towards the male and mate with him, so a flower's colours and stripes, its scent and its nectar, work on the wings of the bees and butterflies and humming-birds. The bees are drawn towards the flowers. Their wings beat and carry the pollen from one flower to {273}
Figure 8.6 DNA with wings: sycamore and dandelion seeds.
another. The wings of bees can truly be called flowers’ wings, for they carry flower genes just as surely as they carry bee genes.
Elephant bodies cannot tell whether they are working to spread elephant DNA or virus DNA, and bees’ wings cannot tell whether they are working to spread bee DNA or flower DNA. As it happens, if we set aside exceptional cases like the bees that are fooled into wasting their time copulating with bee orchids, they are working to spread both. The difference between ‘own DNA and pollen DNA, from the point of view of the bees’ executive machinery, cannot be perceived. Peacocks and bees, flowers and elephants, stand to their own DNA in much the same relation as they stand to the DNA of parasitic viruses that infest them. Virus DNA is a program that says: ‘Duplicate me in a simple and direct way, using the ready-made machinery of host cells.’ Elephant DNA says: ‘Duplicate me in a more complicated and {274} roundabout way that involves, first, building an elephant.’ Flower DNA says: ‘Duplicate me in an even more complicated and roundabout way: first, build a flower and, second, use that flower to manipulate, by indirect influences such as seductive nectar, the wings of a bee (which has already conveniently been built to the specifications of another lot of DNA, the bee's "own" DNA) to carry far and wide the pollen grains inside which are the very same DNA instructions.’ We shall approach this conclusion again, from another direction, in the next chapter. {275}
<<
* * *
>>
CHAPTER 9
WE HAVE JUST CONCLUDED THAT FLOWERS AND elephants are, in effect, hosts to their ‘own’ DNA in the same kind of way as they are hosts to virus DNA. This is correct but it leaves difficult questions unanswered. There is an important step missing from the argument. Computer viruses have an easy time of it because the world is already full of computers, powerful, ready and waiting to obey instructions. But these computers are man-made. They are handed to the parasitic programs on a plate. DNA viruses, too, have their hosts, with all their elaborate instruction-obeying cellular machinery, handed to them on a plate. But in the case of living machinery, where does the machinery come from?
Imagine something like a computer virus which, instead of having a ready-made computer all set up to obey its instructions, had to start from scratch. It couldn't just say ‘Duplicate me’, because there is no computer there to obey the instructions. To be truly self-spreading, in a world without ready-provided computing and duplicating machinery, what would a self-duplicating computer program have to do? It would have to begin by saying: ‘Make the machinery needed to duplicate me.’ And, before that, it would have to say, ‘Make the parts with which to assemble the machinery to duplicate me.’ And before that, it would have to say, ‘Gather the raw materials necessary to make the parts.’ This more elaborate program needs a name. Let's call it the ‘Total Replication of Instructions Program’ or TRIP. {276}
The TRIP has to have control over more than just an ordinary computer with a keyboard and a screen. It has to have at its disposal the equivalent of skilled hands, or grasping and manipulating devices, coupled to sensing devices in order to fashion the parts and cobble them together. The hand-like devices are necessary in order to find and assemble the parts and before that gather their raw materials. A computer can simulate things on its screen, but it cannot, on its own, build another computer like itself. To do that it needs to reach out into the real world and manipulate real, solid metals, silicon and other materials.
Let's look a little closer at the technical problems involved. Modern desktop computers can manipulate coloured shapes on cathode-ray screens, coloured pigments on printer paper and sometimes other things like sounds in stereo loudspeakers. These can all be used to create illusions of three-dimensional
solidity but they really are illusions, reliant upon tricking human brains. A cube is drawn in perspective on the screen. With appropriate surface rendering it looks convincingly solid but you still can't actually pick it up and feel it, solid and weighty, between your finger and thumb. With suitable software, you could simulate cutting the cube in half and view the cross-section on your screen. But again it would not really be solid. Computers of the future may fool other senses in similar ways. Future equivalents of the computer mouse may be rigged to convey to the fingers a realistic feeling of inertia as they push a ‘heavy’ object around the screen. But still the object would not really be heavy, would not be made of tangible, solid stuff.
Our computer that runs the TRIP has got to manipulate more than the human imagination. It has to be capable of handling solid objects out there in the real world. How might a computer do this? It would be formidably difficult. We can begin to see this by trying to design a new kind of a computer printer, a ‘3-D printer’. An ordinary computer printer manipulates ink on a two-dimensional sheet of paper. One way of approximating a three-dimensional representation of, say, a cat's body, would be a set of serial sections printed on transparent sheets. The computer would laboriously slice and scan its way through the cat, from nose to tail, printing out hundreds of sheets of {277} acetate. When the sheets are eventually stacked up into a solid block, a three-dimensional view of the cat would be visible inside the block.
This is still not a true 3-D printer, because the cat, when printed out in this way, would be embedded in a matrix of acetate. We might improve matters if we replaced the ink by a self-hardening resin. The sheets would be stacked up as before, and then dissolved or etched away leaving only the now-hardened resin. In the improbable event that the technical problems with this design could be overcome, we'd have an instrument capable of building up any three-dimensional object: a truly three-dimensional computer printer.
Our 3-D computer printer is still deeply rooted in two-dimensional preconceptions. It contrives its three-dimensional result using the principle of serial sections or slices. No output device that relied upon the serial-slice principle would be adequate for our TRIP. A useful machine, such as an internal-combustion engine, could never be made by the serial-slice technique. It needs sub-components like cylinders and pistons, flywheels and belts. These components are made of different materials from each other, and they have to be free to move relative to each other. The engine cannot be built of stacked-up slices: it must be assembled by bringing together previously manufactured, disjoint parts. The previously manufactured parts themselves will need assembling from smaller parts in the same way. The appropriate kind of output device for the TRIP is not a 3-D printer at all. It is an industrial robot. It has a pincer or some equivalent of a hand, capable of grasping objects. The ‘hand’ must be on the end of an arm-equivalent and it must have a universal joint or a set of joints capable of moving it in all three planes. It has the equivalent of sense organs, capable of guiding it towards the next object that must be picked up, and capable of steering that object towards its desired destination so that it can be fastened in position by an appropriate means.
Industrial robots of this kind do exist in modern factories (Figure 9.1). They do work, provided each one has a very particular task to perform at a particular point in an assembly line. But a normal industrial robot is still not adequate to run the TRIP program. It can put parts together — assemble them — if those parts are handed to it in a {278}
Figure 9.1 Industrial robot from Nissan car factory. Yokohama.
fixed orientation, or regimented past it on a production line. But the whole point of our exercise is to get away from things being handed in fixed orientation, ‘on a plate’. Our robot has somehow to find the raw materials for making the parts before it can begin to assemble them together. In order to do this it has to move around the world, actively seeking raw materials, mining them, gathering them up. It has to have the means to travel — something like caterpillar tracks or legs. There are robots that do have legs, or other means of moving around the world in a quasi-purposeful way. The one in Figure 9.2 happens to be rather insect-like, except that it has four legs instead of six. It is provided with sucker feet like a fly, because its parlour trick is climbing up vertical surfaces. A favourite game of its makers is teasing it by placing a hand in just the place where the robot wants to step. The robots foot senses the consequent unsuitability of the terrain and goes into a delightfully life-like pantomime of searching for {279}
Figure 9.2 Walking robot on sucker legs from Portsmouth Polytechnic, England.
a better surface. But this is a detail of one particular robot. An earlier famous robot, the Machina speculatrix ‘tortoise’ built by W. Grey Walter of Bristol University, used to plug itself into the mains to recharge its batteries. As its batteries ran down, it manifested an increasingly restless ‘appetite’ for electricity and intensified its search for a mains plug. When it found one it backed on to it and stayed there until replenished. These details are not fundamental. We are talking about a machine that is capable of moving around on its own limbs and restlessly searching for something under the control of its own sense organs and its own on-board computer.
Our next task is to join the two kinds of robot together. Imagine that the walking, sucker-footed robot carries, on its back, something like the industrial, hand-wielding robot that we saw earlier. The combined machine is under the control of an on-board computer. The on-board computer has a lot of routine software for controlling the legs and the sucker feet, and for controlling the arm and hand assembly. {280} But it is under the overall control of a master Duplicate Me program which fundamentally says: ‘Walk around the world gathering up the necessary materials to make a duplicate copy of the entire robot. Make a new robot, then feed the same TRIP program into its onboard computer and turn it loose on the world to do the same thing.’ The hypothetical robot that we have now worked towards can be called a TRIP robot.
A TRIP robot such as we are now imagining is a machine of great technical ingenuity and complexity. The principle was discussed by the celebrated Hungarian-American mathematician John von Neumann (one of two candidates for the honoured title of the father of the modern computer — the other was Alan Turing, the young British mathematician who, through his codebreaking genius, may have done more than any other individual on the Allied side to win the Second World War, but who was driven to suicide after the war by judicial persecution, including enforced hormone injections, for his homosexuality). But no von Neumann machine, no self-duplicating TRIP robot, has yet been built. Perhaps it never will be built. Perhaps it is beyond the bounds of practical feasibility.
But what am I talking about? What nonsense to say that a self-duplicating robot has never been built. What on earth do I think that I myself am? Or you? Or a bee or a flower or a kangaroo? What are all of us if not TRIP robots? We are not man-made for the purpose: we have been put together by the processes of embryonic development, under the ultimate direction of naturally selected genes. But what we actually do is exactly what the hypothetical TRIP robot is defined as doing. We roam the world looking for the raw materials needed to assemble the parts needed to maintain ourselves and eventually assemble another robot capable of the same feats. Those raw materials are molecules which we mine from the rich seam of food.
Some people find it offensive to be called a robot. This is usually because they think that a robot has to be a jerky, moronic zombie with no fine control, no intelligence and no flexibility. But these are not necessary or defining properties of a robot. They just happen to be properties of some of the robots that we have built with present-day technology. If I say that a chameleon, or a stick insect or a human {281} is a robot that carries its own programming instructions about inside it, I am not saying anything at all about how intelligent it is. An entity can be very intelligent and still be a robot. Nor am I saying anything about how flexible it is, for a robot can be very flexible. Twentieth-cent
ury people who object to being called robots are objecting to a superficial and irrelevant association of the word (like an eighteenth-century person who objects to calling a steam carriage a vehicle of transport on the grounds that it doesn't involve a horse). A robot is any mechanism, of unspecified complexity and intelligence, which is set up in advance to work towards fulfilling a particular task. The TRIP robot's task is to distribute copies of its own program about the country, together with the machinery necessary to execute the program.
The starting point of our discussion of self-copying robots was this. We decided that a simple Duplicate Me program, like a computer virus or a real DNA virus, was all very well, but it depended upon the world being very cushy — set up with machinery capable of reading and obeying the instructions. But the world is cushy like that only because somebody, or something, else has already built that instruction-obeying machinery. We've now imagined a highly sophisticated robot which is, once again, a gigantic digression on a Duplicate Me program. Instead of just saying ‘Duplicate me’, the program says: Assemble the parts and make a new version of the entire machinery needed to copy me, and then load me into its on-board computer.’