Seminar

Abstract:

Spiking Neural Networks (SNNs) resemble biological information processing mechanisms to exploit their low energy-consumption and complex processing capabilities. In parallel we see powerful new applications rising regularly, with generative AI being the latest trend. In contrast to SNNs, these applications leverage the more abstract principle of artificial neural networks (ANNs). SpiNNaker2 was designed to exploit hybrid mechanisms of SNNs and ANNs as well as symbolic computing to lower the footprint of mainstream AI and enable new applications from edge to cloud scale. We will go from the roots of the SpiNNaker architecture in brain-simulation, over robotic control to state-of-the-art AI such as large-language-models and drug discovery.

Dresden University of Technology, the University of Manchester, Racyics GmbH and Globalfoundries (GF), recently announced the tapeout of the SpiNNaker2 chip, a new artificial intelligence chip inspired by the human brain.