STARBRAIN Project at STARLAB
Things are moving fast at STARLAB's Brain Building Project. On my first day (Feb 7th 2000) at STARLAB in Brussels, Belgium, I was told that STARLAB will create a brain building team of up to 12 people, which includes two full time STARLAB employees (myself and my ATR colleague Dr. Andrzej Buller) and up to 10 students. STARLAB has 150 applicants from MIT for summer studentships, so we will have no shortage of talent.
Dr. Buller suggested a new name for STARLAB's brain building effort, namely "STARBRAIN". This suggestion was approved by STARLAB management, so from now on, the old term "CAM-Brain Project" will be dropped and will be replaced with the term -
STARBRAIN
STARLAB's Artificial Brain Project
We will also be changing another name. The first major application of the artificial brain is intended to be the control of a cute robot kitten that we called "ROBOKONEKO" (Japanese for "robot kitten"). Since the funding for the construction of this kitten robot will have to be found in Europe, a Japanese name for it no longer seems appropriate, so we have anglicized it to "ROBOKITTY".
STARLAB has bought the world's 3rd CAM-Brain Machine (CBM), which will be delivered May/June 2000. The STARBRAIN team will be housed in the nicest and largest room in the STARLAB building (according to the opinion of Walter de Brouwer, STARLAB's CEO), which suggests that STARLAB management sees STARBRAIN as one of its glamor projects. This is encouraging. Certainly there is strong media interest in STARBRAIN. Within the first few days of being here, Discovery Channel and UK's Channel 4's Equinox will start TV documentaries on the project and my ideas, with several other TV documentary crews in the pipeline. Two TV documentaries have already been aired (Dutch and German). Hopefully all this publicity will help Starlab's IPO value when the company goes public in May 2001.
I will collaborate with some local university students (M.Sc. and PhD level) to help me and Dr. Buller design the world's first artificial brain. This will be an immense job. The CBM can handle 64000 evolved circuit modules, interconnected to form the brain. Each of these modules needs to have its function specified, and its role in the larger architecture defined. We will need all the manpower we can get. Once the CBM is here, we can rapidly explore the evolvabilities and MECs (Modular Evolvable Capacities) of individual modules to see what they can do. Once we get a feel for this, we will then start building multimodule systems, with 10s of modules, then 100s, 1000s, to the limit of 64,000.
We also need to be thinking about the design of the second generation CBM, which will probably handle a million modules, with a billion neurons. We will probably need some kind of methodology to evolve the interconnections BETWEEN modules. Thinking along these lines has barely started, but its obvious no human team can easily design a million module system BY HAND. The process will have to be automated somehow. Ideas on how to do this are welcome. Progress in electronics (Moore's law) is racing ahead of our ability to dream up applications for its growing and massive potential.