Anderson on Chaos and Evolution

Phil Anderson is an interesting, combative figure and also one of the greatest living physicists. (His faculty page, which looks like rank self-promotion, is actually pretty accurate.) I'll write more on his running duel with Steven Weinberg about reductionism -- for now, I'd just like to flag this passage from his paper on evolution and quenched chaos (Proceedings of the National Academy of Sciences 80, 3386 (1983)) as a description of what nonreductionist arguments are supposed to do:

I hope that this model is capable of mimicking the behavior of the origin of molecular evolution, in the sense that a modern-day statistical physicist would define as "being in the same universality class with the origin of life." That is, we cannot hope by the finiteness of our lives to work in the original time scale, nor can we guess precisely the chemical nature of the actual molecules or the boundary conditions and constraints which were present. What we can do is attempt to show that in a well-defined mathematical model, which in principle contains no inherent fudge factors that prejudice the outcome, a transition such as that between inanimate molecules and life does occur. [...] In particular, "chance and necessity" alone will not do it, even aided by the idea of self-organization via dissipative structures. I argue that, in addition, chaos is a precondition...

Statistical mechanics is a little like modal logic; you prove (usually without much rigor) that a feature X, e.g. life or superconductivity, is common to all worlds that meet a criterion Y. Criterion Y is usually something like "matter consists of particulate bits," which is (eventually) a fact from particle physics. The Anderson-Weinberg dispute comes down to whether it makes sense to say therefore that particle physics "proves" X -- or, in Weinberg's phrase, whether the "arrows of explanation" for Y lead to particle physics. The real world at the Planck scale leads uniquely to one of the gazillion possible worlds at the atomic scale that would lead to the observed world at the everyday scale. If you knew the Final Theory, you could (in principle) plug it into an infinitely powerful computer, solve for infinitely many elementary particles with the appropriate initial conditions, and discover superconductivity or life. In practice, however, science proceeds backwards, and you generally don't need to go far back to find something that would count as an explanation.