Speaker of Workshop 2
Will talk about: Principles of neocortical self-construction
Rodney Douglas is Professor Emeritus at the Institute of Neuroinformatics of the Swiss Federal Institute and the University of Zurich. He graduated in Medicine and Neuroscience at the University of Cape Town before moving to the MRC Anatomical Neuropharmacology Unit in Oxford to research the anatomy and biophysics of the neuronal circuitry of cerebral cortex together with Kevan Martin. As Visiting Associate, and then Visiting Professor at Caltech, he extended his interests in neural computation by simulation of neocortical circuits (together with Christof Koch), and also by their emulation as neuromorphic electronic systems (together with Misha Mahowald and Carver Mead). In 1996 he and Kevan Martin established the Institute of Neuroinformatics in Zurich. In 2000 Douglas was awarded the Koerber Foundation Prize for European Science; and in 2008 he was elected to the Swiss Academy of Technology and Sciences. Douglas's research interests include: experimental anatomy and physiology of visual cerebral cortex; theoretical analysis and simulation of cortical circuits and their development; and the design and fabrication of hybrid CMOS VLSI neuromorphic systems. He is a co-founder of the neurotechnology spinoff 'iniLabs'.
Current scientific wisdom in Europe and the USA promotes exhaustive data collection projects as the necessary route to understanding the structure and function of the nervous system, and so of future neuromophic computers. These proposed exa- to zettabyte descriptions stand in stark contrast to the gigabyte of construction information available to the developing brain. This enormous disparity raises the question of how the elaborate information processing circuits of (for example) the neocortex construct themselves using the relatively small amount of information encoded in the genome of neuronal stem cells. Our approach to this intriguing question combines experimental observation of cortical development with simulation of a detailed model of the physical process itself. The entire simulated development plays out under the control of an abstract regulatory network inserted into the initial neuroepithelial cells. These cells then expand by mitosis, differentiation, and morphological specialization into the multi-layered connected neural networks of two example murine neocortical areas, which are composed of about 0.25M neurons. I will explain this process, and show how we are able to infer the control GRN from only sparse experimental data. I will argue that understanding such abstract principles of biological development can provide novel insights into brain organization and function, as well as offering novel approaches to future self-constructing computers and other manufacturing technologies.