12/2/02

1. This is a short sketch of my argument that there wasn't enough time for evolution to do what we see has been done.

2. In short, the theory presents a plausible scenario for the evolution of the biological "hardware" but it does not adequately explain the generation of the biological "software". To expand on Paley a little, suppose instead of finding a watch, you found a CD all loaded up with a PC operating system. In my opinion, evolutionary theory has a plausible explanation for how the plastic disc evolved by chance mutations and natural selection, first by certain molecules polymerizing, and then,...(many technical details here)... voila, you get a nice shimmering CD. That's the easy part. The hard part is how did the bit pattern get constructed?

3. I think you can draw a parallel with software development here. A typical genome, if I'm not too terribly wrong, contains about the same order of magnitude of information (bit count) as a typical computer operating system. We humans know what it takes (what it takes us anyway) to build an operating system that will not only boot up, but run a computer effectively. What it takes is a lot of thought and development. The program is not written one bit at a time, nor could it even be understood that way. Instead, there are high level concepts developed long before the first bit is written down. These concepts are fleshed out, distributed out to more or less independent developers where they are further developed. My point is that there is an enormous amount of symbolic depiction of abstract concepts which all must make sense at every one of many levels. The final result is a vast collection of bits that come together to work together in a pre-planned, conceptually structured, way. And, they usually don't work as planned. So, there are many iterations of testing, failing, diagnosing, repairing, and even re-designing, before the thing can be used on a computer.

4. So, armed with that knowledge (of how we build operating systems), we can imagine trying to build an operating system without the high level conceptualization and without the many types of symbolic representation of the final product. Instead, we would ask developers to induce some random changes into some of their code, try it, see if it works at all, if it does, see if it will work with everything else in the operating system, if so, send out a version with those changes and see how it flies in the market place. We can imagine automating this process so that the "mutated" code gets automatically generated at the rate of, say, one per millisecond. The testing too would be automated, so that only successful "mutations" would drop out the other side to be integrated with various versions of the rest of the operating system. We might expect a "good" mutation to drop out every few weeks or so. The integration process would also be automated so that different combinations of "mutated" code from all the different components would be tried together. We might expect that once a year we would see an integrated operating system that would even run. But how long would it take to get all the bells and whistles of a modern operating system to appear using this kind of development technique? I think it would take more time than the universe has existed.

5. But to top it off, the job of an operating system is vastly simpler than that of a genome. The operating system only has to run a specific piece of hardware. The genome not only has to run a specific piece of hardware, but it must fix that hardware when it breaks, it must make all its parts, not only for repair but for the original construction as well, it must make the factories which make the parts, it must make eggs or sperm cells and endow them with sufficient information to make another similar (not identical) piece of hardware that can do all these same things. Makes the job of building operating systems seem like a piece of cake.

6. So, it seems to me that if we could get some plausible estimates on how many iterations it would take to get functional operating systems using the non-planned, non-designed development technique I described, and compare that number of iterations to the number of iterations available to organisms in 4 billion years according to evolutionary theory, my bet would be that there is not enough time. And that for a vastly harder problem to boot.