This is a special post for quick takes by leogao. Only they can create top-level comments. Comments here also appear on the Quick Takes page and All Posts page.
The modularity assumption: Model computation can be explained in terms of a relatively small number of modules, each of which is a clean abstraction over its internals. Each module can also be recursively explained in terms of submodules, all the way down to the atomic computations of the model. There are no irreducibly complex interactions that can only be modeled by keeping a huge number of modules in mind at once.
i want to draw an analogy between modularity in models and biology, where biology is also designed by a blind optimization process.
question: to what extent is biology evidence for the modularity assumption? on the one hand, organs are modules; whereas you could have imagined some vastly more complicated world where everything is just one complicated mass of hopelessly entangled bullshit. on the other hand, metabolic pathways seem ridiculously complicated, and it often seems like everything is mangled up with everything else in a complicated way.
I suspect the key distinction here is not exactly modular vs. non-modular, but "learnable in small chunks" vs. "not learnable in small chunks". If there is a way for evolution/Adam to learn connections between modules such that each individual connection term learned improves the free energy, those connections will be learned, because each individual connection is a small chunk and thus quick to find by random guessing. If there is a way to learn more and more of these connections until the modules cease to be separate at all, that will happen as well. But... (read more)
Try KEGG, they have maps on the front page, just click the links.
1Q Home
Without context it doesn't really tell much, because even some simple things can be drawn as complex diagrams. But perhaps the only way to learn context is to do serious study...