Simultaneous Multi-Module Evolution: The Need for Simultaneous Evolution of Multiple Neural Net Modules for Brain Building
Hugo de GARIS, Jian Tang, Di Huang · 2007
This paper discusses a problem that our brain building research group is confronted with increasingly each year, namely, How can we evolve, one neural net module at a time, while building artificial brains, when the technology, i.e. Moore's Law, keeps increasing the number of neural net modules that a PC can process in real time, by a factor of two nearly every year or two? In 2006, it was possible to have a PC process the neural signaling in real time, of an artificial brain of some 50,000 modules, each to be evolved individually, by a human BA (Brain Architect) or EE (evolutionary engineer). This number is already too large to be practical, so the obvious question arises, How to automate the simultaneous evolution of multiple numbers of neural net modules? This paper presents the problem and gives some tentative answers.