October 12, 2023. Thursday. 3PM. TG23 (Town Hall).
Speaker: Roman Belavkin (Middlesex University, London)
Title: Information Processing in the Human Brain: What are Our Models Missing?
Abstract: Artificial Intelligence (AI) has been evolving using two approaches: Symbolic and Sub-symbolic (connectionist). Artificial Neural Networks (ANNs) are examples of the latter, and recent successes of Deep Learning have made the term “AI” almost synonymous with deep neural networks. Super-human performance has been achieved in many areas including image recognition, computer and strategic games (e.g. chess, Go), and the progress in text processing by Large Language Models (e.g. Chat GPT, Bard) has even prompted some to suggest that General AI is within reach. In this talk, I will remind some of the long-standing problems and philosophical questions that were posed in the early days of AI and that remain unanswered. Then I will compare some numerical data on human nervous system to estimate and compare its information processing capacity with that of modern AI applications. Then I will remind about important properties of biological nervous systems that our even most modern ANNs ignore. I will argue that although modern ANNs can be comparable to the human brain in terms of the number of information processing units, the way this information is processed by the human brain could be by orders of magnitude more superior to that of our current ANNs. Developing the necessary mathematics and algorithms will certainly improve our models, but many questions of AI, especially of more philosophical nature, will remain unanswered.
Bio: Roman Belavkin obtained MSc in Physics from the Moscow State University and PhD in Computer Science from the University of Nottingham. His research interests span several areas including geometric analysis of optimal and learning systems, dynamics of information, value of information, quantum information, topology of information, geometry and combinatorics of mutation and recombination of sequences, optimal control of evolutionary algorithms, cognitive modelling. Roman joined Middlesex University in 2002, where he participated in several research projects and organized research seminars of the Artificial Intelligence group. From 2009 Roman has been the Principle Investigator of the EPSRC project “SANDPIT: Evolution as an Information Dynamic System”, which was led by Middlesex University in collaboration with Universities of Manchester, Keele and Warwick. In this project, Roman developed a theory of optimal control of mutation rate in evolutionary systems, and the team discovered plastic mutation rates in microbes. Roman’s current work is on geometric and dynamic value of information theory, which has applications in parameter control and optimization of learning, adaptive and evolving systems. Roman has many international collaborations: He has been an associate member of the “Centre of Applied Optimization” in the University of Florida, USA; his collaboration with Tokyo University of Science was recognized in 2014 by the award from the university’s president Professor Akira Fujishima. Roman has been a keynote speaker at many international conferences, workshops and research seminars. He also serves on the editorial board of the “Optimization Letters” and “SN Operations Research Forum” journals.