AI Virtual Rat by Harvard and DeepMind Sheds Light on Brain-Body Connection

No robot has come close to matching the evolutionary wonder of human and animal agility. Harvard neuroscientists have developed a virtual rat with an artificial brain that can walk like a real rodent to shed light on the enigma of how brains govern movement.

Professor Bence Ölveczky of the Department of Organismic and Evolutionary Biology oversaw a team of researchers that worked with experts at Google’s DeepMind AI lab to create a digital model of a rat that is biomechanically accurate.

Also Read: China Creates World’s First 3D e-Skin, Helping Robots Feel Human-like Sensing

They taught an artificial neural network, or the virtual rat’s “brain,” to manage the virtual body in a physics simulator called MuJoco, where gravity and other forces are present, using high-resolution data recorded from real rats.

According to Oňlveczky, an expert in conditioning real rats to learn complex behaviours so that their neural circuitry can be studied, the researchers discovered, publishing in Nature, that activations in the virtual control network accurately predicted neural activity measured from the brains of real rats producing the same behaviours.

According to Ölveczky, the achievement is a novel method for examining how the brain regulates movement since it uses developments in deep reinforcement learning, artificial intelligence, and 3D movement monitoring in freely moving animals.s.

The cooperation was “amazing,” according to Oňlveczky.”DeepMind created a pipeline to teach biomechanical agents how to navigate intricate environments.”We lacked the necessary resources to train these networks through such simulations.”

Also Read: AI Robot Makes History: Climbs China’s Great Wall

Similarly, co-author and Google DeepMind Senior Director of Research Matthew Botvinick described working with Harvard academics as “a really exciting opportunity for us.” The difficulty of creating embodied agents—AI systems that must think intelligently and convert that thought into actual physical behaviour in a complex environment—has taught us a great deal. 

It was likely that applying a similar strategy in the field of neuroscience may help shed light on how the brain and behaviour are related.

Although the focus of Ölveczky’s lab is on basic issues about brain function, the platform might be utilized, for instance, to develop more advanced robotic control systems. Giving the virtual animal autonomy to complete tasks similar to those faced by actual rats could be the next stage. 

“We have many ideas about how these tasks are solved and how the learning algorithms that underpin the acquisition of skilled behaviours are implemented based on our experiments,” Ölveczky said.”We hope to test these theories and further our knowledge of how complex behaviour is produced in real brains by utilizing the virtual rats to begin with.”

Also Read: AI Steve: A businessman proposes the first AI MP Candidate for Brighton in the UK General Election