|
|
|
|
MkI was Lucy's first physical body. It was a pile of junk, fundamentally. Many robots cost thousands or even hundreds of thousands of pounds, but Lucy 1 cost no more than a weekend break in Bognor Regis.
Nevertheless, Lucy 1 has managed to demonstrate some important principles and fly in the face of much received wisdom. The objective was, and still is, to find a single neural architecture that is capable of self-organising into a wide range of virtual machines, capable of the manifold tasks involved in seeing, hearing, thinking and moving. By the end of Phase 1 of the project her brain consisted of a number of "modules" (cortical maps), each of which performed a different function. Each map was unique, but nevertheless they all had enough in common to offer exciting hints about a universal architecture. Lucy 1's body had a lot of limitations. Her vision was very poor and her movements restricted. So in 2002 we started work on Lucy MkII - a rather more ambitious but powerful robot. We expect Lucy to "grow up" in this way several times over the coming years. A comprehensive overview of the development of Lucy MkI and the ideas underlying the project can be found in this book. While you're waiting for your copy to arrive, click on the link bar for more information. |
Copyright © 2004 Cyberlife Research Ltd. |
The
mechanisms in Lucy 1's brain were also very crude in several ways. Nevertheless,
her neural network was extremely large by conventional standards. It was
composed of over 50,000 virtual neurons, or more accurately neural columns, each
of which is made
from a small circuit of neurons. Some of the structures in her brain were biologically
very plausible, while others were much more speculative and exploratory.