Paul Muhle-Karbe

I investigate the brain mechanisms that enable rapid adaptation to novel tasks. Humans are very proficient at solving novel problems, due to their capacity to generalise prior knowledge from related past experience. For instance, when traveling abroad, we can quickly work out how to use an unknown train ticket machine. This ability to transfer knowledge flexibly between tasks is a key building block of adaptive behaviour, and also presents an unsolved problem for current forms of artificial intelligence.

I want to understand the brain mechanisms that allow humans to form these flexible mental programs that can be redeployed under different circumstances. By combining different brain recording techniques with pattern classification methods, and computational models, I characterise the representations that form as humans learn to perform novel categorisation tasks.

This work aims to reveal fundamental biological mechanisms that underpin goal-directed behaviour in complex, open-ended environments. In the long run, it also aims to provide new insights that could be of help in building more intelligent machines.