Since the beginning of the computer age, the terms "information processing'' and "computing'' have been used to describe the dynamics of natural adaptive systems, ranging from the brain to the immune system, cellular metabolism, and genetic regulation. It is widely assumed that information processing in such systems takes place in a very different manner than in our current day computers. In particular, the architectural features of these natural systems—large and varying numbers of relatively simple, stochastic, and noisy components, limited, dynamic, and unreliable connections, and no central control—require a radically different model of information processing than the traditional one.
In this talk I describe the mechanisms underlying information processing in biological systems and discuss the relevance of these mechanisms for adaptive behavior in these systems. I then abstract four general principles that I claim are key to adaptive information processing in decentralized systems such as these. These principles deal with the representation and transmission of information, the essential role of randomness, the importance of fine-grained parallel architectures, and the interplay of bottom-up and top-down processes in all such systems. Finally, I describe how these principles might be used in artificial intelligence applications to achieve robust and fluid pattern recognition and learning.
Melanie Mitchell received a Ph.D. in Computer Science from the University of Michigan in 1990. Since then she has held faculty or professional positions at the University of Michigan, the Santa Fe Institute, Los Alamos National Laboratory, the OGI School of Science and Engineering, and Portland State University. Her research interests include Intelligent systems, machine learning, complex systems, evolutionary computation and artificial life. Professor Mitchell holds the 21st Century Research Award from the J.S. McDonnell Foundation, a five year grant for research in complex systems. She seeks to understand how natural systems perform computation, and how to use ideas from natural systems to develop new kinds of computational systems. Professor Mitchell also has interests in Cognitive science, particularly computer modeling of perception and analogy-making, emergent computation and representation, and philosophical foundations of cognitive science.
David Maier