I am creating a simulation to evolve simple, multi-cell organisms. (Just for fun!)
Neural networks are fairly easy to evolve, even more so when it’s done by random mutations and not actual training. Build an ANN at random and introduce mutations with every generation. The ANNs that accomplish simple goals (by pure chance) live to duplicate with every evolutionary cycle. Fairly easy stuff.
I am stumped when it comes to creating something that would simulate the genes that represent a body. After some reading today, there isn’t much info on how cells form into specific shapes for arms, hands, organs, etc. (I am sure there is a ton of data, but I don’t know what subject to Google.)
Genes can create the patterns for specific chemicals and cells. How to cells then develop into functional body parts? What makes a heart the shape of a heart?
I think that having a better understanding of that concept can help me develop a framework for physical evolution, even if it as a very tiny scale.
(Putting the ANN in charge of controlling those different body parts is also easy. It’s just a matter of allowing those physical traits to evolve first.)
I would just add that it’s a gradient of transcription factor(s), proteins that regulate production of other proteins. Some of the regulated proteins are other thranscription factors that create finer gradients, like repeated stripes of gradients. along the initial, single gradient. This can be repeated on multiple levels so that pretty much every point of the body has unique combination of transcription factors. This combination triggers all the other proteins, including structural that create the form of the body. and of course, all this doesn’t just happen in 3D, but also changes with time. It’s complicated.
For illustration and possible direction for some particular interesting cases wrapped in a very entertaining package, I recommend this: https://www.youtube.com/watch?v=ydqReeTV_vk