Using genetic algorithms to simulate evolution of structure


IIUC the logic behind the direction of Numenta is to arrive at AGI by reverse-engineering an existing system – the biological brain.

If we are to use nature as inspiration, wouldn’t it make sense to go the whole distance?

The Homo Sapiens brain is the product of the biological organism shaped by the forces of evolution.

How about modelling a basic braincell, figuring out intelligent degrees of freedom (dendrite tree density, dendrite length, dendrite:axon ratio, # synapses, plasticity falloff with time, etc.) and create an artificial environment with a reward system & allow the artificial organisms to fight it out?

Would this not be an interesting experiment? We may find that the resultant structure bears some similarities to what we observe.

^ This video highlights the amazing results that can be achieved using genetic algorithms. Respect to Karl Sims for doing this before the days of Unity3D.


Would this require lots of processing power?


The first challenge is to figure out if it is a viable line of research, and if anybody is doing it.

Processing load can only be estimated once we have a fairly clear line of sight solution path.

It would be misdirection to worry about that.

1 Like