I added in an edit a reference as to how immune system, basically, operates. You have population of b-cells, which evolves for elimination of foreign substances. Good ol evolution re-used to evolve a part of the b-cell genome, inside your body. The results seem very impressive—recognition of substances—but all the heavy lifting is done using very simple and very stupid methods. If anything, our proneness to seasonal cold and flu is a great demonstration of the extreme stupidity of the immune system. The viruses only need to modify some entirely non-functional proteins to have to be recognized afresh. That’s because there is no pattern recognition going on what so ever, only incredibly stupid process of evolution of b-cells.
If I was trying to claim that immune systems were complex in a way that is similar in nature to learned cortical algorithms then I would be thoroughly dissuaded by now.
The immune system is actually a rather good example of what sort of mechanisms you can expect to evolve over many billions generations, and in which way they can be called ‘complex’.
My original point was that much of evolutionary cognitive science is explaining way more complex mechanisms (with a lot of hidden complexity. For very outrageous example consider preference for specific details of mate body shape, which is a task with immense hidden complexity) as evolving in thousandth the generations count of the immune system. Instead of being generated in some way by operation of the brain, in the context whereby other brain areas are only marginally less effective at the tasks—suggesting not the hardwiring of algorithms of any kind but minor tweaks to the properties of the network which slightly improve the network’s efficiency after the network learns the specific task.
We probably don’t disagree too much on the core issue here by the way. Compared to an arbitrary reference class that is somewhat meaningful I tend to be far more likely to accepting of the ‘blank slate’ capabilities of the brain. The way it just learns how to build models of reality from visual input is amazing. It’s particularly fascinating to see areas in the brain that are consistent across (nearly) all people that turn out not to be hardwired after all. Except in as much as they happen to be always connected to the same stuff and usually develop in the same way!
I added in an edit a reference as to how immune system, basically, operates. You have population of b-cells, which evolves for elimination of foreign substances. Good ol evolution re-used to evolve a part of the b-cell genome, inside your body. The results seem very impressive—recognition of substances—but all the heavy lifting is done using very simple and very stupid methods. If anything, our proneness to seasonal cold and flu is a great demonstration of the extreme stupidity of the immune system. The viruses only need to modify some entirely non-functional proteins to have to be recognized afresh. That’s because there is no pattern recognition going on what so ever, only incredibly stupid process of evolution of b-cells.
If I was trying to claim that immune systems were complex in a way that is similar in nature to learned cortical algorithms then I would be thoroughly dissuaded by now.
The immune system is actually a rather good example of what sort of mechanisms you can expect to evolve over many billions generations, and in which way they can be called ‘complex’.
My original point was that much of evolutionary cognitive science is explaining way more complex mechanisms (with a lot of hidden complexity. For very outrageous example consider preference for specific details of mate body shape, which is a task with immense hidden complexity) as evolving in thousandth the generations count of the immune system. Instead of being generated in some way by operation of the brain, in the context whereby other brain areas are only marginally less effective at the tasks—suggesting not the hardwiring of algorithms of any kind but minor tweaks to the properties of the network which slightly improve the network’s efficiency after the network learns the specific task.
We probably don’t disagree too much on the core issue here by the way. Compared to an arbitrary reference class that is somewhat meaningful I tend to be far more likely to accepting of the ‘blank slate’ capabilities of the brain. The way it just learns how to build models of reality from visual input is amazing. It’s particularly fascinating to see areas in the brain that are consistent across (nearly) all people that turn out not to be hardwired after all. Except in as much as they happen to be always connected to the same stuff and usually develop in the same way!