When: Thursday, December 8, 2022, 11:00 AM - 12:00 PMWhere: Zoom
Speaker: Mark Sandler, Professor of Signal Processing in the School of Electronic Engine
Title: Artificial Neuroscience: using Neuroscience to inspire advances in the understanding of artificial Neural Networks, leading to explainable & controllable, engineered & simplified Artificial Intelligence.Zoom link: https://qmul-ac-uk.zoom.us/j/81148100921Abstract: It is several months since I first used the phrase ‘Artificial Psychology’ as an attempt to capture some of my bewilderment and dissatisfaction at the state of play in Deep Learning research. It is curious how a single term, a new name, drove me forward to try to re-evaluate Deep Learning and its catalogue of techniques and jargon. Over those months, I have formulated and re-formulated my thoughts and conceived the term ‘Artificial Neuroscience’ to accompany Artificial Psychology and expand the analogies that each suggests.
In this talk, I will share some of the motivation behind these labels and will hint at how I imagine a spruced-up Deep Learning methodology should look. Central to this is the notion that Computer Science should no more monopolise Artificial Intelligence than Biology monopolises Neuroscience. Rather Deep Learning needs to evolve and in doing so, include the methodologies and learned practices of several other disciplines. The one I focus on in this talk is Linear Algebra, but I will also touch on how, as it continues to develop, DL needs to bring in knowledge and learning from Psychology, Electronic Design, Behavioural Sciences, several other branches of mathematics, Cognitive Development, Law and Ethics. And probably several more. I hope the presentation will be provocative, and generate discussion and thought among us.