de Garis - from guest editor
of an academic journal to associate editor
For the past 6 months (Q4 2000 - Q1 2001), on and off, I've been busy guest editing a special issue on "Evolutionary Neural Systems" for the academic journal Neurocomputing, published by Elsevier. The editor in chief of this journal, Dr. David Sanchez must have liked the work I did, since he has recently invited me to become an associate editor of this journal. This appointment became official in mid May 2001. I'm already on the editorial board of the journal "Evolutionary Computation", so this official recognition is gratifying and a kind of completion, since the main thrust of my work is the evolution of neural networks. I think I was the first to evolve neural networks with dynamic output signals (in 1989).
Also, I feel the need to solidify the more traditional side of my work to act as a counterweight to my far more controversial side, namely sticking my neck out trying to build an artificial brain, and especially my major claims that a gigadeath war is coming late 21st century over the issue of species dominance - the whole artilect/terran/cosmist thing. The more conservative, non visionary, of my colleagues sneer at my attempts in this regard, so I need to mollify them to some extent by having solid academic credentials.
I suppose the next step up would be to become an editor in chief of a journal, but that would require huge amounts of work, which would only lower the probability of me succeeding in building the world's first artificial brain. However, I would consider it if the journal were on "Brain Building", but the field has not yet been established. I have still not yet "proved concept" that it is possible. Right now I still need to show that the neural net circuits that I grow and evolve in special hardware at electronic speeds have sufficient "evolvabilities" to be useful. It is possible that the neural net model implemented in the electronics is not sufficient, not sophisticated enough for good useful evolvabilities. My team keeps trying to evolve interesting neural circuit modules, but we may eventually have to upgrade the model itself, or worse, have to design a 2nd generation machine, with more modern chips to achieve the greater evolvabilities.
In the longer term, this is not a big deal, since Moore's law will allow a new brain building machine generation, and its corresponding artificial brain, every 4-5 years Im reckoning. I hope to be involved personally in creating 5 of those generations before I retire in my 70s. By 2020, Moore's law will give us "Avogadro Machines" with a trillion trillion components - molecular engineering. We will have all the tools we need to build true artificial brains.
In short, this brain building business is taking time. It's not easy. The French media are starting to ask the question "where's the brain?" and fair enough. Its coming, its coming. 4 CBMs now exist in the world. We're working on the individual modules. Once we feel the evolvabilities are good enough, we will move up to multi-module architectures, 10s of modules, then 100s, 1000s, 10,000s up to 64000, the capacity of the CBM. If we can prove concept, then sufficient numbers of other researchers will get interested, so that a new journal, specifically dedicated to "Brain Building", can be established. I would be proud to be editor in chief of such a journal, despite the enormous amounts of time it would eat up.