• Week 3 (May 10th) - Professor Mike Wheeler (University of Stirling): Perception, Action, and the Extended Mind
• Week 4 (May 17th) - Professor Zoltan Dienes (University of Sussex): Using Bayes To Get The Most Out Of Null Results
• Week 5 (May 24th) - Professor J. Kevin O'Regan (Director, Laboratoire Psychologie de la Perception, Institut Paris Descartes de Neurosciences et Cognition): Why red doesn't sound like a bell: the sensorimotor approach, with computational and empirical applications
• Week 6 (May 31st) - Professor Owen Holland with Dr Bhargav Mitra and Dr David Devereux (University of Sussex): Right body, right mind?
• Extra Talk (July 19th - Out of term) - Professor Geoffrey Hinton (University of Toronto): How to force unsupervised neural networks to discover the right representation of images

Week 3

Date: May 10th
Speaker: Professor Mike Wheeler

Abstract: Perception, Action, and the Extended Mind

According to the extended cognition hypothesis (henceforth ExC), there are actual (in-this-world) cases in which thinking and thoughts (more precisely, the material vehicles that realize thinking and thoughts) are spatially distributed over brain, body and world, in such a way that the external (beyond-the-skin) factors concerned are rightly accorded cognitive status (where 'cognitive status' signals whatever status it is that we ordinarily grant the brain in orthodox, non-extended cognitive theory). David Chalmers, one of the original architects of ExC, has recently articulated an objection to the view which turns on the claim that the idea of cognitive extension is in conflict with an intuitive thought that we ought to preserve. Chalmers puts that intuitive thought like this: 'It is natural to hold that perception is the interface where the world affects the mind, and that action is the interface where the mind affects the world. If so, it is tempting to hold that what precedes perception and what follows action is not truly mental.' Chalmers proceeds to offer a defence of ExC against the worry. In my talk I'll (i) set the scene with some comments about how one ought to understand ExC (comments that involve some criticisms of Andy Clark's version of the view), (ii) explain Chalmers' objection and his response to it, (iii) argue that Chalmers' response fails, and (iv) suggest that we should solve the problem by ditching the intuitive thought. This final move will enable me to address a challenge that, up until now, has arguably not been met successfully by advocates of ExC, that is, to say what consequences the view has for empirical work in cognitive science and psychology.

Speaker Bio

Mike Wheeler is Professor of Philosophy at the University of Stirling. Prior to joining Stirling Philosophy in 2004, he held teaching and research posts at the Universities of Dundee Oxford, and Stirling (a previous appointment). His doctoral work was carried out in the School of Cognitive and Computing Sciences at the University of Sussex. His primary research interests are in philosophy of science (especially cognitive science, psychology, biology, artificial intelligence and artificial life) and philosophy of mind. He also works on Heidegger and is interested in exploring ideas at the interface between the analytic and the contemporary European traditions in philosophy. His book, Reconstructing the Cognitive World: the Next Step, was published by MIT Press in 2005.

Week 4

Date: May 17th
Speaker: Professor Zoltan Dienes

Abstract: Using Bayes To Get The Most Out Of Null Results

Users of orthodox statistics, including psychologists, have typically not interpreted null results in a principled way, resulting in mistaken conclusions and choices of research direction based on non-significant findings. A non-significant result does not distinguish no evidence for an effect from evidence for no effect, radically different states of affairs. The distinction falls out naturally from a Bayesian analysis, in a consistent way that cannot be accomplished with orthodox statistics, even when statistical power is taken into account. I will show how simple free online software can be used to determine what a null result is actually telling us. I also comment on how Bayesian analysis frees us from three logical and moral paradoxes that orthodox statistics forces on us (to do with planned versus post hoc tests, stopping rules, and multiple testing). A paper discussing all these issues is here: http://www.lifesci.sussex.ac.uk/home/Zoltan_Dienes/Dienes%202011%20Bayes.pdf

Speaker Bio

Zoltan Dienes is professor of psychology at the University of Sussex. His main research interest is the distinction between conscious and unconscious mental states, especially in the context of implicit learning and hypnosis. He is the author of Dienes, Z. (2008). Understanding Psychology as a Science: An Introduction to Scientific and Statistical Inference. (Palgrave Macmillan)

Week 5

Date: May 24th
Speaker: Professor J. Kevin O'Regan

Abstract: Why red doesn't sound like a bell: the sensorimotor approach, with computational and empirical applications.

Why does red look red, rather than looking green, or rather than sounding like a bell? Indeed why does red have a feel at all? Why do pains hurt instead of just provoking avoidance reactions?

The "sensorimotor" approach provides a way of answering these questions by appealing to the idea that feels like red and pain should not be considered as things that are generated in the brain, but rather as things that we do. I shall show how this a priori counter-intuitive idea gives rise to successful computational models and/or empirical predictions about space and color perception and sensory substitution. In addition to helping understand human consciousness, the approach has applications in virtual reality and in robotics.

Speaker Bio

After studying theoretical physics at Sussex and Cambridge Universities, Kevin O'Regan moved to Paris in 1975 to work in experimental psychology at the Centre National de Recherche Scientifique. Following his Ph. D. on eye movements in reading he showed the existence of an optimal position for the eye to fixate in words. His interest in the problem of the perceived stability of the visual world led him to question established notions of the nature of visual perception, and to discover, with collaborators, the phenomenon of "change blindness". His current work involves exploring the empirical consequences of a new "sensorimotor" approach to vision and sensation in general. He is particularly interested in the problem of the nature of phenomenal consciousness, which he addresses experimentally in relation to sensory substitution and pain, and theoretically in relation to space and color perception. He is interested in applying this work to robotics. Kevin O'Regan is currently director of the Laboratoire Psychologie de la Perception, CNRS, Université Paris Descartes.

Week 6

Date: May 31st
Speaker: Professor Owen Holland with Dr Bhargav Mitra and Dr David Devereux

Abstract: Right body, right mind?

Most research robots, including humanoids used for research in cognition, are designed to be easy to control, and are therefore conventionally engineered, with simplified joints enabling precise movements. However, the increasing concern with the relevance of embodiment to cognition has led us to question this approach. In order to discover the extent to which a specifically human embodiment may enable and constrain specifically human cognitive abilities, we are building a humanoid which copies the human musculo-skeletal embodiment, with an emphasis on the elastic properties of muscles and tendons. Our project has three aims: to build such a robot, to find out how it can be controlled, and to investigate whether the embodiment does in fact lead to specifically human cognitive characteristics. In this talk we will describe and discuss our progress towards achieving the first two aims, and outline possible approaches to the third.

Speaker Bios

Owen Holland is Professor of Cognitive Robotics at Sussex. After studying psychology at Nottingham, and lecturing in it at Edinburgh, he spent some years in commerce and in the electronics industry. He became interested in robotics in the late 1980s, and was a founder member of the Intelligent Autonomous Systems Laboratory at UWE Bristol (now the Bristol Robotics Laboratory) in 1993; he has since worked at the University of Bielefeld, the California Institute of Technology, the University of Essex, and a number of private research laboratories. He came to Sussex in 2009. His research is mainly biologically inspired (swarm robotics, swarm intelligence, energetically autonomous robots, machine consciousness, and cognitive robotics), and he is also interested in the history of cybernetics.

Dr Bhargav Mitra is a Postdoctoral Research Fellow in Computer Vision. He obtained a BTech in Engineering from the University of Kalyani, West Bengal, in 2002, and a DPhil in Engineering from the School of Engineering and Design, University of Sussex, in 2010.

Dr David Devereux is a Postdoctoral Research Fellow in Robotics. He obtained an MEng in Computer Science and Cybernetics from the University of Reading in 2005, and a PhD in Robotics from the University of Manchester in 2010.

Out of term

Date: July 19th
Speaker: Professor Geoffrey Hinton

Professor Geoffrey Hinton will be receving an honarary degree at the Sussex University Graduation Ceremony on the 20th July, and has kindly agreed to give a public COGS Seminar Series talk at the University on the 19th.

Abstract: How to force unsupervised neural networks to discover the right representation of images

One appealing way to design an object recognition system is to define objects recursively in terms of their parts and the required spatial relationships between the parts and the whole. These relationships can be represented by the coordinate transformation between an intrinsic frame of reference embedded in the part and an intrinsic frame embedded in the whole. This transformation is unaffected by the viewpoint so this form of knowledge about the shape of an object is viewpoint invariant. A natural way for a neural network to implement this knowledge is by using a matrix of weights to represent each part-whole relationship and a vector of neural activities to represent the pose of each part or whole relative to the viewer. The pose of the whole can then be predicted from the poses of the parts and, if the predictions agree, the whole is present. This leads to neural networks that can recognize objects over a wide range of viewpoints using neural activities that are "equivariant" rather than invariant: as the viewpoint varies the neural activities all vary even though the knowledge is viewpoint-invariant. The "capsules" that implement the lowest-level parts in the shape hierarchy need to extract explicit pose parameters from pixel intensities and these pose parameters need to have the right form to allow coordinate transformations to be implemented by matrix multiplies. These capsules are quite easy to learn from pairs of transformed images if the neural net has direct, non-visual access to the transformations, as it would if it controlled them. (Joint work with Sida Wang and Alex Krizhevsky.)

Speaker Bio

Geoffrey Hinton received his BA in experimental psychology from Cambridge in 1970 and his PhD in Artificial Intelligence from Edinburgh in 1978.  He did postdoctoral work at Sussex University and the University of California San Diego and spent five years as a faculty member in the Computer Science department at Carnegie-Mellon University. He then became a fellow of the Canadian Institute for Advanced Research and moved to the Department of Computer Science at the University of Toronto. He spent three years from 1998 until 2001 setting up the Gatsby Computational Neuroscience Unit at University College London and then returned to the University of Toronto where he is a University Professor. He is the director of the program on "Neural Computation and Adaptive Perception" which is funded by the Canadian Institute for Advanced Research.

Geoffrey Hinton is a fellow of the Royal Society, the Royal Society of Canada, and the Association for the Advancement of Artificial Intelligence. He is an honorary foreign member of the American Academy of Arts and Sciences, and a former president of the Cognitive Science Society. He received an honorary doctorate from the University of Edinburgh in 2001. He was awarded the first David E. Rumelhart prize (2001), the IJCAI award for research excellence (2005), the IEEE Neural Network Pioneer award (1998) and the ITAC/NSERC award for contributions to information technology (1992) and the NSERC Herzberg Medal which is Canada's top award in Science and Engineering.

Geoffrey Hinton investigates ways of using neural networks for learning, memory, perception and symbol processing and has over 200 publications in these areas. He was one of the researchers who introduced the back-propagation algorithm that has been widely used for practical applications. His other contributions to neural network research include Boltzmann machines, distributed representations, time-delay neural nets, mixtures of experts, Variational learning, products of experts and deep belief nets.    His current main interest is in unsupervised learning procedures for multi-layer neural networks with rich sensory input.

Series organized by Prof Steve Torrance stevet@sussex.ac.uk