Back
Biological neural systems have always fascinated and inspired AI researchers. By scaling the models of biological neural networks from 1950s, we have reached today’s AI, which is impressive, but far from universal. Today’s AI has a number of fundamental limitations in latency, efficiency, and tasks it can solve. These limitations are particularly devastating in robotics. To overcome the limitations, we look back to biological neural systems. In this talk, I will show how new bio-inspired neural models with stateful, adaptive, and branching neurons and dynamic, plastic, and sparse connections lead to AI systems that can be integrated in real-time control loops, can learn continually, and run on board on an embedded computing device. This new bio-inspired AI may lead to more reactive, safe, efficient, and capable robotic assistants.
Yulia Sandamirskaya (Zurich University of Applied Sciences (ZHAW))
Yulia Sandamirskaya is a full professor at Zurich University of Applied Sciences (ZHAW), leading a Research Centre Cognitive Computing in Life Sciences in Wädenswil and a Neuromorphic Computing research group. She obtained a Diplom in Physics with honours from the Belarussian State University and a Dr. rer. nat. degree from the Institute of Neural Computation (summa cum laude) at the Ruhr-University Bochum. She was leading a research group at the Institute of Neuroinformatics at UZH/ETH Zurich, as well as the Applications Research team of the Neuromorphic Computing Lab at Intel. She coordinated an EU CSA project NEUROTECH: Neuromorphic Computing Technology and was listed among 30 women in robotics you need to know about by robohub in 2020 and among ''Female Future Makers'' by Business insider in 2022.
More information