Comment by uplifter

> how to square the idea that evolution produces knowledge with the idea that it doesn't optimize even for reproductive fitness

Its really fairly simply: natural selection requires two things: heritable genetics and a source of variation in the genetics between individuals. Mutation is the most basic source of variation, and that produces new information. But new information isn't necessarily knowledge. Assuming a scientific testing gloss, each new genetic code variation X can be considered as a hypothesis, that "variant X is fit", and then natural selection events that act on copies of X (for or against) serve as experiments testing the hypothesis. Through iterative experiments, we weed out the copies of the variants where the hypothesis of them being fit was proved by natural selection to be false, and what remains should be those copies of genetic variants which have (mostly) proven to be true. Learning and understanding which variants are fit (where the hypotheses are true) is knowledge, and in this way evolution produces knowledge while not having any optimization goal (in the intent sense, which I agree is a requirement for something to be meaningfully "optimizing" anything, because you can't aim in a direction without a sense for that direction).