The field of Artificial Intelligence (AI) has made impressive progress over the last couple of years: voice-controlled digital assistants, automated translation services, self-driving cars, world-class Go playing algorithms and other novelties have made it into homes and headlines. Most of these advances are based on a method called `Deep Learning’ and – together with many successes – the limitations of this technology have come fully to light: Compared to human intelligence, AI today still needs enormous amounts of training data and energy – and it can be easily fooled.
The human brain seems to recognize the world in a fundamentally different way: learning mostly in an unsupervised fashion with minimal amounts of labeled training data. While Deep Learning is based on neuroscientific models from the mid of last century, we believe that the time has come for a new generation of machine learning algorithms which are inspired by 21st-century brain research. We approach this challenge with an interdisciplinary team of mathematicians, computer scientists, and neuroscientists, who aim to rethink the foundations of artificial and biological intelligence.