Science

Understanding the Brain to Improve Artificial Intelligence

This text is part of a dedicated Research section.

A recent study published in Nature Communications sheds light on how the brain learns. Advances that can help improve AI performance.

“Our starting point is that all cognitive processes that take place in our brains, including learning, are physical processes, so they can be modeled and analyzed using computers,” says Eilif B. Muller, assistant professor of neuroscience at the University of Montreal. The researcher has been working on brain modeling for twenty years, using mathematics and physics to better understand how the brain learns. “Learning is one of the most interesting processes: it is where our adaptability comes from,” he adds.

It was in Switzerland, as part of the international Blue Brain Project team, that Professor Müller began to explore the neocortex. This initiative was launched by Professor Henry Markram of the Swiss Federal Institute of Technology in Lausanne in 2005 to model the neocortex, that region of the brain that resembles a walnut shell. It is on this surface with a thickness of 2 to 3 mm that adaptive learning takes place: speech development, understanding of mathematics, learning from errors, processing of images, sounds, smells, etc.

virtual simulation

Brain circuits are made up of neurons connected by synapses. For several years, scientists in the field believed that what we learn is recorded or stored in our brains as a long-term change in the connections between neurons across synapses. This synaptic plasticity will be the substratum of learning.

Eilif B. Müller and his collaborators at the Blue Brain project have done physical modeling of learning processes in the neocortex. “We have been trying for years to find rules, patterns for these changes. It’s like a few puzzle pieces, some of which are missing, and there’s no reference image,” Muller says. To do this, they built a virtual replica of the neocortex of a young rat, which they trained using data and then compared their predictions with reality.

The results, published in Nature Communications, are amazing. “We have found a single rule that combines the experimental observations we have right now and our model can make predictions,” Muller says. Now the team hopes to repeat this experience.

At the intersection of neuroscience and AI

This brain research, although fundamental, opens the way to several practical applications, helping to understand the mechanisms of neurodevelopmental disorders such as autism, schizophrenia or Down syndrome, among other things. But a better understanding of this important area of ​​the brain could also help create better artificial intelligence. “AI researchers are actively trying to build systems that harness the powerful learning abilities of the neocortex,” Muller notes.

Montreal is also an important environment for this research, at the interface between neuroscience and artificial intelligence, with the presence of Yoshua Bengio (Scientific Director of IVADO) and his colleagues, several organizations and support important to governments. That is why Professor Muller decided to come.

“The goal of our research is twofold,” says Muller: as we better understand the mechanisms of learning, we can implement them in new approaches to artificial intelligence. But AI’s mathematical language can also help neuroscience: “AI has developed mathematical concepts and languages ​​that can describe the rules for learning in artificial systems,” says Muller, who seeks to bring the two fields of inquiry together.

Artificial intelligence… more human

But what will “smarter” AI do? At the moment, AIs are trained immediately, using huge amounts of data, so that they learn how to make classifications (to distinguish, for example, a dog from a cat). But once all this training has been done, the machine will stop adjusting (if, for example, we notice that a chihuahua has been tagged as a cat and not a dog). The algorithm cannot do this continuous learning and will forget everything it already knows when new data is presented to it.

The way our synapses respond could inspire new ways to tackle this and other problems. “We see big differences between how algorithms learn and what we know about synapse changes in the brain. This is a contradiction that fascinates scientists, and this is a very fertile ground for cross-fertilization of ideas,” concludes Müller.

See in video

Back to top button

Adblock Detected

Please consider supporting us by disabling your ad blocker.