Wolf De Wulf

Hi, my name is Wolf De Wulf and I am a computer scientist from Brussels.
I studied artificial intelligence (AI) at the Vrije Universiteit Brussel, and I am currently researching AI in biomedical contexts, as part of the 2022 cohort of the UKRI CDT in Biomedical AI at the University of Edinburgh.

I am generally interested in how circuits of neurons interact to form intelligence, I believe it a fundamental aspect of both neuroscience and artificial intelligence (AI) research. On the one hand, understanding the neural codes behind a particular behaviour facilitates the development of treatments for conditions related to that behaviour. On the other, it can yield inspiration towards novel AI models that mimic the brain. At the University of Edinburgh I get to work with researchers from the School of Informatics and from the Centre for Discovery Brain Sciences, resulting in an interdisciplinary approach to research that I genuinely enjoy.

I am currently pursuing a PhD in computational neuroscience, jointly supervised by Professor Matthias Hennig and Professor Matthew Nolan. We are trying to understand the neural substrate for spatial navigation, which is generally thought to be handled in the hippocampal formation.

We look specifically at the medial entorhinal cortex (MEC). The MEC has traditionally been thought of as containing various types of functionally defined neurons, whose activity correlates with variables such as position, speed, head direction, and more. In recent years, this model of the MEC has been amassing counter-evidence. It seems that the activity of MEC neurons is much more heterogeneous than previously thought, with findings of neurons correlated to various mixes of these positional variables and surprising differences across tasks. These findings suggest that we should re-evaluate the way we think of the neural spatial code in the MEC:

• Are functionally discrete neuron types the building blocks of the cognitive map?
• Are they an epiphenomenon of global circuit computation?
• Is it reasonable to link neuron activity to human-intuitive variables such as position and speed?
• Should we stop trying to interpret single neuron activity and look at the population code to understand computation?

These are just a subset of the many questions I am thinking about in my current project.