Neuroscience and artificial intelligence, an interdisciplinary project

Miguel Aguilera has been awarded one of the twenty-five competitive La Caixa – Postdoctoral Junior Leader Fellowship – Incoming grants

Researcher Miguel Aguilera has been selected with the prestigious La Caixa Postdoctoral Junior Leader Fellowship – Incoming 2023 program. Twenty-five new La Caixa Found Incoming Fellows have been selected from all the applications. This fellowship provides forty researchers, 25 incoming and 15 retaining fellowships, with the opportunity to further develop their scientific careers and create innovative and ground-breaking science in Spain and Portugal.

After several years of postdoctoral fellowships, Miguel Aguilera joined the Basque Center for Applied Mathematics - BCAM in 2022 as an Ikerbasque Research Fellow in the Mathematical, Computational and Experimental Neuroscience (MCEN) research line. In Aguilera's words, "for the first time, I have the possibility of building a research plan with longer-term objectives, and the Junior Leader Fellowships are a great opportunity for young researchers to move in that direction".

Miguel Aguilera's project aims to explore the connection between transformer models – the technology that is behind the recent groundbreaking advances in large language models like chat GPT – and neuroscience – in particular, the question of how hippocampal circuits encode experience-dependent spacial information. "In both cases, it is very challenging to study how (artificial or biological) neural networks behave in interaction with dynamic changes in sensory streams and internal states," points out Aguilera. Using methods from nonequilibrium statistical physics, he will develop simplified mathematical descriptions of the ongoing interaction of neural networks and their continuously changing environments.

This research will help to better understand models of artificial intelligence (AI) and biological neural circuits. The methods Aguilera will use can help provide a deeper understanding of experimental observations recorded in laboratories through high-throughput data acquisition tecnologies, able to capture the activity of thousands of neurons in behaving animals. "For AI, in the long term, finding mathematical approximations of the behaviour of large neural networks could help reduce the huge energy consumption and carbon footprint involved in training and running models with billions of parameters", explains researcher Miguel Aguilera. Finally, finding connections between how artificial and biological networks adaptively process information can provide new insights into how the brain and the human mind work.

 Fig.1 Comparison between symmetric (left) and asymmetric (right) Hopfield networks and transformer models.

The La Caixa grant offers a stimulating, cutting-edge research environment for researchers, as well as the opportunity to work closely with some of the leading experts in their field. Miguel Aguilera will collaborate closely with researchers in the MCEN research group, which is staffed by experts in algebraic topology, "I hope that by analysing data from hippocampal circuits we can use this knowledge to establish connections between the topological properties of neural network models and the properties of the space through which the animal navigates," says the BCAM researcher.

"I share common interests with BCAM members on topics ranging from non-equilibrium statistical physics to neural network inference methods in machine learning that are directly related to this project and could lead to exciting collaborations. I also hope to collaborate with researchers from other Basque centres such as Biofisika or Biocruces," adds Aguilera. The grant recipients have access to the network of leading research centres in Spain, and work on projects spanning a wide range of disciplines, from biotechnology to physics.

The La Caixa Postdoctoral Fellow Junior Leader is one of the prestigious grants offered by the La Caixa Foundation and is designed to attract the most talented and promising researchers from around the world to work in Spain. Miguel Aguilera applied with an interdisciplinary project. "When I applied, I was worried that the project might be too interdisciplinary to receive a good evaluation," he says. After the evaluation, the panel praised this aspect of the project positively, "this is not always the case, but given the positive feedback, I would advise applicants not to be afraid to explore bold and creative ideas in their proposals. Of course, there is some risk in doing so, but I think more and more assessment panels understand better the value of transdisciplinary science. As candidates, I believe we also have the power to slowly change the system and contribute to making science exciting and curiosity-driven", concludes Aguilera.