Tom Pittman's WebLog

(or something like that)

2008 July 16 -- Artificial Life

One of the things that keeps me coming back for more Michael Crichton novels is that they tend to have good science. Prey is not one of those. There are two kinds of science described in this 2002 novel, and Crichton bungled both.

First, this is about computers -- very tiny computers, but Crichton is at pains to show us some of their programming. The code he shows is not in any common industrial programming language like Lisp (which might be appropriate for the artificial intelligence of this story) or C++; perhaps his computer consultant invented it, or maybe it's a conflation of several older languages.

I know something about computers. My degree is in computer science. Recall also that the weakest link in Jurassic Park was his description of the computer operations; if it had not been for less than credible descriptions of computer bugs, I would have had trouble believing the dinosaurs never happened. Prey has serious computational flaws from the beginning. My guess is that few people understand the technology well enough to notice the flaws, so that they have no personal basis for rejecting it. And those who do understand it, they are so enthusiastic in their agreement with Crichton on the other scientific flaw as to give this one a pass.

Consider memory. Crichton belabors the memory issue, so this is not unfair. He has his nanobot computers consuming the memory chips out of electronic gadgets, but leaving the CPU chips alone. Any electrical engineer will tell you that there is no fundamental difference between memory chips and processing chips, they are both made from exactly the same technology and processing steps. Two transistors arranged in series or parallel form a gate (decision logic), two gates cross-coupled form a memory cell. Really dense memory is made from half a transistor, which holds an electrical charge where the missing part would go, plus the gates to put it there and read it out. Except for the printed part number on the outside package, there is no way to tell these apart. You can't even tell by looking at the circuits inside unless you know a lot about microcircuit technology. Modern CPU chips have thousands or millions of memory cells included for easy access, and memory chips have decision logic built in to refresh those sagging electrical charges. There is no difference at all.

If, as his story has it, the fundamental construction of these nanobot computers is organic, then the computer logic and memory would also be organic, not silicon, so there would be nothing of value in electronic gadgets for the nanobots to "eat", not even for memory. Besides, there is no "memory" substance in chips for them to eat, just silicon with a few impurities, plus some metal connections and glass (oxidized silicon) insulators.

In his story (this is a spoiler) the nanobots learn how to collectively assume human shapes and to move around indistinguishable from the particular humans in the story, matching even the printed pattern on one guy's tee-shirt. Yet, the processing power and memory of the individual nanobots is very small. Necessarily so, if they are each only one tenth the diameter of a human hair. Mimicking humans is a popular concept in fiction, but nothing like that kind of mimicry happens in real life, and for good reason. Generalized color matching happens often, but not the patterns, not the details of other beings. Why is that? There is a huge amount of processing and memory required to do that. Image processing programs are notorious memory and computation hogs. Each surface nanobot might have the capacity to be a spot of arbitrary color, but telling millions of individual nanobots what color it must assume for the total appearance to be correct requires substantial computer graphics processing and communication ability. Even granting the evolutionary hypothesis of the story (I'm coming to that), the complexity of organization is beyond the time frame to support self-organizing.

In case you had not yet guessed, evolution is the other scientific flaw here. Given that he considers evolution central to his story, Crichton should at least know his Darwinian dogma better. Here's his central point, clearly spelled out:

Considering that we live in an era of evolutionary everything -- evolutionary biology, evolutionary medicine, evolutionary ecology, evolutionary psychology, evolutionary economics, evolutionary computing -- it was surprising how rarely people thought in evolutionary terms. It was a human blind spot. [page 325]
This is an accurate observation. What's missing is the reason for it. Crichton novels usually explain not only the observation, but also why it is so. Here, like most Darwinists, he lets his dogma get in the way of science. Darwinism is the established religion of the academic elite, so calling your research "evolutionary" generates Federal grant money, but the reason people don't think in evolutionary terms is that evolutionary ideas do not accurately describe how the world works. Even the biologists who insist that evolution is the only way to understand their science, for the most part they cannot show how evolution significantly impacts what they themselves are doing. You could do exactly the same science from a creationist perspective, and mostly get exactly the same results. Science actually started out creationist, and there still are many closet creationist scientists out there today whose hidden beliefs are not found out because it makes no difference at all. Unless they happen to say something, and then they get fired on prejudice alone and not for any flaws in their science.

Darwinistic evolution is supposed to work by tiny random mutations chancing upon some trait that is beneficial to the survival of the organism, which is then spread throughout the entire population by Natural Selection. Each generation has some number of mutants (initially one) and the remainder unmutated individuals, but the mutants survive better and have more offspring, so there are more of them in the next generation, until all the non-mutants disappear. The literature suggests that it takes something like a thousand or more generations for a good mutation to completely replace the inferior earlier trait. The important factor here is that the beneficial trait must go out into the environment and survive before it reproduces. That way the lower survival rates fail to reproduce and get replaced by the better survivors.

In Prey, the nanobots are manufactured in a giant container -- not exactly a cauldron, but something like that -- by biological "assemblers" (bacteria genetically programmed to do this) which add on the molecules and parts one at a time, thousands and millions in parallel, so they turn out a whole batch in a few hours. That part is credible. Not today (yet) but credible. Of course I'm not a microbiologist. Somehow these assemblers get released into the wild, and the nanobots start to evolve.

Except we have a problem: the nanobots who are out there in the environment experiencing survival (or the lack thereof), they are not the ones doing the fabrication. They have no impact on the fabrication at all, because that is done by assembler bacteria, not nanobots at all. The assemblers can only operate in their cauldron broth, they are not part of the nanobot "swarm" flying around outside and surviving or dying.

It gets worse. Instead of taking thousands of generations to evolve new traits and abilities, Crichton's nanobot swarms are learning and evolving new "emergent" behaviors every generation, every few hours. No honest Darwinist would claim that of their theory. It takes millions of years, they tell us. That's why, apart from a few trivial cases that stop when they reach their designed limits, we cannot see any biological evolution happening in the real world at all.

This being a computational nanobot, Crichton is careful to tell us about (software) "genetic programs" which "evolve" in their computers to do surprising things and learn new behaviors. He even mentions one of the researchers by name, Danny Hillis. Yes, there is such a person, and he did publish a paper on co-evolution, where two separate entitities, a neural net designed to sort 16 numbers, and another "parasitic" net designed to interfere with that sort, were both randomly altered over time, and rewarded by giving the successful mutations (correct or poor sorting, as the case may be) control in the next generation. After some 5000 generations it finally achieved a near-optimal sort in 65 compares (the best possible is 60).

Notice that there is no such thing as self-organization here. These programs were designed to sort (or inhibit it), and they were designed to modify only the choices of what to compare next. There is no anything-goes evolution here, the program would crash instantly if they tried it. And the researchers all know it. Real-life evolution is like that. Improvements are so hard to find, that you really have to know what you are doing, and carefully design the system to allow only safe mutations. And even then the evolution does not improve beyond what the researcher planned for it. Incidentally, real (biological) evolution works the same way.

The physics in this novel is also faulty -- kind of like the TV show MacGyver, plausible if you don't know more than C+ high-school physics and don't try to add up 2+2. These tiny nanobots are so small that they easily float around in the air. Credible. They "maneuver by climbing the viscosity of the air." Less so. Newtonian physics says that the nanobot can only go up by pushing an equal weight of air down (or a smaller weight faster). That's how birds and airplanes and rockets work. Everything else -- including birds and airplanes and rockets -- go down by the pull of gravity. Newton figured that part out too. Small stuff, with more surface area than mass, goes down much more slowly, but it does go down. Let's say these nanobots "climb the viscosity of the air" by pushing air molecules down. That doesn't exactly match Crichton's description, but hey, it's fiction. So they fly around.

You can't see a single microscopic nanobot without a microscope, but a million of them fairly close together could be visible. A million of them touching each other would fill a space smaller than a match head. But they could not fly if they were touching each other, they each need access to motionless air to push down. In the story these nanobots swarm like fish and birds and bees, a visible dark cloud the size of a person. That's far more than a million nanobots, perhaps more like a billion or ten billion to weigh "three pounds." They still need a lot of air space between them to keep from getting blown down by the next unit's prop wash.

Let's ignore the scale problems. By and by the swarms of nanobots get smart and they start to imitate human forms. But they have no substance, the swarm still being mostly air. Towards the end the swarms start getting more massive. Not bigger, just heavier, up to 50 pounds. Here's where the physics breaks. 50 pounds of solid material, no matter how small the individual particles, requires 50 pounds of air moving down (accelerating) at 32 feet per second per second, just to keep from falling out of the air. But these guys still fly. No way. Do you know how big a helicopter rotor must be, and how fast it must move to get that much air moving down? There was no prop wash in Crichton's story. Like I said, it's fiction.

The concern Crichton raises about artificial life taking on its own existence is valid -- if you believe in evolution. But even so, the opportunities for human-made micro-organisms (the worry-warts call it "grey goo") evolving faster than humans can figure out how to eradicate them is infinitesmal. Based on the complexities here, we won't have any opportunity to face the problem for a very long time. As someone at one think tank put it, "The physicists don't always understand biology as well as they think they do, and vice versa." That does not improve by bringing physicists and biologists together in think tanks, the fields themselves are growing too fast.

There is a much bigger and more valid concern over scientists starting with an existing organism with an existing complex reproductive cycle already solved (by God :-) and ignorantly modifying it to do some surprising and dangerous thing, and then that modified organism getting out to wreak havoc on the ecology and/or people. No evolution is needed, and the technology to do it already exists. Fortunately, Natural Selection works in our favor, to eliminate such mishaps. I call it anti-evolution.

Rest of this year's Blog
Complete Blog Index