jacob_cannell comments on Something's Wrong - Less Wrong
You are viewing a comment permalink. View the original post to see all comments and the full post content.
You are viewing a comment permalink. View the original post to see all comments and the full post content.
Comments (161)
I agree with much of what you say, yet . .
Yes, and I was pointing out that this applies equally to biologists acting as amateur engineers.
Your statement seemed to me to be a blanket substance-less dismissal of the original discussion on why the retina's design may not be as suboptimal as it appears to amateur engineers.
I doubt your certainty. Optimality is well understood and well defined in math and comptuer science, and evolutionary algorithms can easily produce optimal solutions for well defined problems given sufficient time & space. Optimality in biology is necessarily a fuzzy concept - the fitness function is quite complex.
Nonetheless, parallel evolution gives us an idea of how evolution can reliably produce designs that roughly fill or populate optimums in the fitness landscape. The exact designs are never exactly the same, but this is probably more a result of the fuziness of the optimum region in the different but similar fitness landscapes than a failure of evolution.
I think this is a mischaracterization of evolutionary algorithms - they are actually extremely robust against getting stuck in local optimums. This is in fact their main claim to fame, their advantage vs simpler search approaches.
You somewhat overinterpret, and also remember that the quote is my summarization of someone else's point. Nonetheless, I stand by the general form of the statement.
It is extremely difficult to say that a particular adaptation is suboptimal unless you can actually prove it by improving the 'design' through genetic engineering.
Given what we currently know, it is wise to have priors such that by default one assumes that perceived suboptimal designs in organisms are more likely a result of our own ignorance.
wnoise answers this for me below, and shows the validity of the prior I advocate