Â鶹´«Ã½ÉçÇøÈë¿Ú

be.AI is a Leverhulme Scholarship Programme in biomimetic embodied Artificial Intelligence. It is an interdisciplinary programme looking at all aspects how intelligence arises from the interaction of the brain with a body and the environment and how the gained insights can be used to build different AI.

be.AI builds on our research strengths at the Â鶹´«Ã½ÉçÇøÈë¿Ú along a number of central research themes.

be.AI is rooted in the understanding gained from biological systems - Animals sense their environment through a number of sensory modalities, including vision, audition, olfaction/taste and tactile sensation. Researchers at Sussex are experts in Sensory Neuroscience, investigating how sensory processing works in animal models, which will inform our understanding of the input space for biomimetic AI. For example, Maravall investigates feature processing in touch perception in the rodent whisker systems, Lagnado, Baden and Niven information processing and transmission in the retina, Nowotny the processing of olfactory cues in insects’ sensilla, and George and Ildiko Kemenes as well as Niven  how memories are formed and maintained. At higher levels of organisation, Eldridge investigates the ecological role of sound in ecosystems.

be.AI is biomimetic - The insights from biological brains lead us to developing novel solutions for AI and autonomous systems. In the EPSRC funded Brains on Board programme grant and ActiveAI international centre-to-centre collaboration Nowotny, Philippides and Graham are investigating (with UK and Australian partners), how insects learn rapidly and perform robust behaviours in complex environments. The resulting computational brain models are distilled into controllers for autonomous robots. One core technology enabling this research is the GeNN software developed by Nowotny that makes use of modern GPU accelerators for brain simulations in the form of spiking neural networks. This is a key technology needed for biomimetic AI systems where spiking neural networks are seen as the third generation of AI. This technology is unique to Sussex and will give students in be.AI a head-start over other institutions.

be.AI is embodied - Financed by the BBSRC, Buckley in Informatics and Lagnado In the Life Sciences investigate the closed-loop dynamics of brain-body-environment interactions in larval zebrafish, using whole brain recordings and Buckley and Niven work with a closed-loop ant VR system to decipher information processing in behaving ants. These experiments critically inform Buckley’s work with Seth on predictive coding that investigates how perception is an active process in the brain and leads to new algorithms for active AI. This relates back to work on active AI by Philippides, Graham and Nowotny (see above) and is complemented by work of Berthouze who, along with collaborators at University College London, investigates fundamental mechanisms underpinning the development of coordinated motor control in both health and disease (in particular Parkinson's Disease). Some of these mechanisms have been tested in developmental robotic systems.

be.AI informs robotics and hybrid systems - The research on embodiment in biological systems is complemented by work from Husbands, Philippides and Buckley on evolutionary robotics and how complex behaviours can emerge through the combination of simple controllers and multiple feedback loops with an agent’s body and the environment. This includes, for instance, the emergence of walking behaviours in legged robots with decentralized controllers which are able to exploit chaotic dynamics to produce highly resilient embodied systems. Nowotny also works on hybrid systems, connecting brains to brain-models in a “digital twin approach” to better understand brain function through computational mimicry.

Consciousness in be.AI - Within the Sackler Centre for Consciousness Science, Seth, Roseboom and Barrett investigate a broad range of questions relating to consciousness: how to measure it, how it arises and how it interacts with and enables intelligence. This leads to research on what role the equivalent of consciousness has to play in AI. Seth is collaborating on this with members of CIFAR in Canada, including Yoshua Bengio, a leading figure in AI.

be.AI needs to be ethical - Some of the biggest issues in classical deep learning and machine learning relate to its lack of human interpretability, fairness in decision making and uncertainty quantification. Within Sussex, Quadrianto, Simpson and Sharmanska work on improving upon more traditional AI systems for understanding and generating tabular, photographic, medical imaging, and video data, to account for these important human considerations.

Using be.AI methods to understand human behaviour - Roggen is interested in understanding human behaviour utilizing lifelong learning approaches inspired by adaptive capabilities of biological organisms, to address “open-ended” scenarios which tend to be of high societal value. In Psychology John Drury investigates social behaviour via crowd simulations and experiments.

Chatwin, Young and Birch are applying AI in multi-camera tracking of people and objects while in NLP,  Weeds, Weir and Carroll  study how meaning is created and shared via language, combining the now common-place data-driven methods with the need to understand how humans manage more complex nuances with less data and little learning time.

be.AI and music - Eldridge, Kiefer and Magnusson all work with dynamical systems, AI and programming as part of their musical practice. As founding members of the and members of the Sussex Humanities Lab, they apply AI in musical context, ranging from embodied musical interfaces to computational creativity, as exemplified in their .

Philosophy of be.AI – Cognitive philosophers Clark and Chrisley investigate key conceptual issues concerning the nature of mind and intelligence, the differences between biological intelligence and current AI systems, and the cognitive impact of socio-technological scaffolding. Clark’s work on embodiment, cognitive extension, and the predictive brain provides a bridge between core concerns within AI and robotics, work on cognitive prosthetics and designer environments, and foundational questions about the nature of mind and the origins of conscious awareness.

How to Apply | Apply Now | Suggested Supervisors | Project Ideas | People in be.AI