Turning human intelligence into AI

Artificial intelligence (AI) is the ability for machines to learn from experiences or information and mimic how the human mind works. For example, to solve problems and make decisions. AI is already part of our everyday lives, powering the voice and facial recognition features on your smartphone. AI is transforming many industries including healthcare, where it’s helping us to develop drugs faster, perform surgery and even diagnose medical conditions.

These advances are thanks to ever-faster hardware, access to huge amounts of data for training, and increasingly complex algorithms — sets of rules machines follow to solve problems or run calculations. But AI still has a major limitation: its ability to use what it’s learned in a new context.

Key to improving this could be understanding how our brains are so good at processing imperfect information. For example, how does a child learn so quickly to recognize a cat when each cat looks slightly different, and there are countless other variables like viewing distance, angles and lighting? Why is it so effortless for us?

Our new collaboration with the Transylvanian Institute of Neuroscience (TINS) in Romania will explore this question by conducting research on the fundamental principles of intelligence.
 

Our brains are made up of billions of neurons connected by trillions of synapses to form an incredibly complex repeating pattern. One objective of the research is to develop a theory for how that network and changes in brain activity bring about that moment of understanding. In other words, the moment you know you’re looking at a cat.

It’s thought that the human brain uses a ‘master’ algorithm for solving seemingly very different tasks – from recognizing a familiar face to moving around an obstacle — and that this is stored in the neocortical microcircuit, a tiny but crucial component of the brain. Researchers will analyze nerve activity in the neocortex to see how this influences the brain’s computing power. This should lead to a better understanding of the master algorithm which underlies higher intelligence.

“If there is one key structure in the brain that holds the answer to intelligence, it’s probably the neocortical microcircuit,” says Raul C. Muresan, President and Principal Investigator. “By investigating the structural and dynamic properties of this building block, we will attempt to extract the fundamental principles of the neocortical fabric and translate them into novel computational models.”

The experienced neuroscientists from TINS will join forces on this project with our AI Research Team, which explores neuroscience-inspired approaches to developing novel AI.

“A deeper understanding of the brain’s computational mechanism will be key to overcoming many of the current limitations in existing artificial intelligence technologies,” says Helmut Linde, our Global Head of Data Science. “Just as mathematicians have developed ‘probability theory,’ on which statistics are based, we need a ‘theory of intelligence’ with which we can develop new algorithms.”

Read more in our press release here.