Don Batten posted a document Genetic algorithms-do they show that evolution works? which includes a critique of the Ev model. On this page I will put some responses. They will be limited for now because I have other things to work on (ie real publications). Readers are encouraged to identify his errors and confusions. It won't be hard, but listing all of them will be a big task!

• "Generation time is ignored. A generation can happen in a computer in microseconds whereas even the best bacteria take about 20 minutes. Multicellular organisms have far longer generation times."

  The (incredible!) unstated implication is that by going slower the simulation somehow fails. However, it is easy to slow down a simulation, and obviously one would get the same results! Another (incredible!) unstated implication here is that the earth is not 4 billion years old. It took 2 billion years for visible life forms to appear. Obviously these had to be pretty sophisticated by that point. So the entire objection holds no water.

• Batten's objections are:

  "However, many of the same problems outlined above also apply to this programming exercise. For example, the selection coefficient is extremely high, the genome is extremely small, the mutation rate high, no possibility of extinction is permitted, etc."

  • the selection coefficient is extremely high. Why does a dandelion produce so many seeds? One plant could repopulate the entire yard! Why does a man or a sea urchin make millions of sperm? The answer is easy: most don't survive. But this objection is again a distraction from the point of the simulation, which was to demonstrate generation of information. Since does that, the objection is mute.
  • the mutation rate high. Perhaps Batten did not read the paper carefully. In Ev the mutation rate is only an order of magnitude higher than HIV. So instead of the simulation taking 700 generations it would take 7000. At 20 minutes per generation, this is only 4 hours. We have 2 to 3 billion years available.
  • the genome is extremely small. This is not relevant to the problem. If you make the genome larger it only slows down the simulation (you are welcome to try it yourself) but the same results will be obtained. If you don't believe that claim, do the simulation yourself.
  • no possibility of extinction is permitted. This is not relevant to the problem of information gain since extinction would merely stop the simulation. Besides, the simulation was explicitly designed to model gain of information in a new genetic control system that the organism is not yet dependent on. Logically, any new genetic control system has this property! So any extinctions that occur at the start are irrelevant. The relative survival will cause the evolution of the sites. If you don't believe this, try a simulation with extinction allowed.

• "Note that we are not saying that mutations and natural selection cannot generate information (see Spetner's book, Not by Chance for example).

  Wow! They are backing off!! They admit that the process can generate information!

  It's just that with real world generation times, real-world sized genomes and real-world organisms which have to survive through multi-dimensional adaptive traits, there has not been enough time to generate even a tiny amount of the biological information seen in living things."

  Clearly a "tiny" amount can be generated, as this is amply demonstrated by the Ev program. Furthermore multidimensionality is important, but notice that there is independence of traits (they are called 'genes'), so selection can proceeed on one gene while merely maintaining others. A billion years is a long time ...

Schneider Lab
origin: 2001 Nov 26
updated: 2002 Jan 22