MIT researchers claim to have solved the puzzle behind the interactions of two neurons, which will unlock a new class of speedy artificial intelligence (AI) algorithms. It’s part of a growing branch of research in which studying the brain is helping create advanced new forms of AI. “Brain research aims to understand the communication between individual neurons or groups of neurons and can help us understand natural intelligence,” Rahul Panat, a professor of mechanical engineering at Carnegie Mellon University, who was not involved in the MIT research, told Lifewire in an email interview. “Natural intelligence can then be used to develop artificial intelligence.”
Thinking Better
The MIT team created a neural network that outperformed state-of-the-art counterparts on a slew of tasks, with considerably higher speedups and performance in recognizing human activities from motion sensors, modeling the physical dynamics of a simulated walker robot, according to a new paper in Nature Machine Intelligence. For example, the new models were 220 times faster on a sampling of 8,000 patients on a medical prediction task. “The new machine-learning models we call ‘CfC’s’ replace the differential equation defining the computation of the neuron with a closed form approximation, preserving the beautiful properties of liquid networks without the need for numerical integration,” MIT professor Daniela Rus, the senior author on the new paper, said in the news release. “CfC models are causal, compact, explainable, and efficient to train and predict. They open the way to trustworthy machine learning for safety-critical applications.” Some brain-inspired computer advances are already available. Reinforcement learning (RL) is a popular family of AI algorithms inspired by recent advances in brain research that have been used in self-driving cars and robotics, Vasileios Christopoulos, a professor of bioengineering at the University of California, Riverside, said in an email interview. “RL is an adaptive process in which species utilize their previous experience in order to make recent advances in understanding the neural mechanisms underpinning RL helped engineers to develop novel AI algorithms in which autonomous agents learn to behave in unknown environments by performing actions and monitoring the consequences of their actions,” he added.
Pondering the Future
Panat predicted that increasing recording capabilities will eventually allow communication between different parts of the brain to be interpreted and deciphered by scientists. “The nervous system that controls peculiar behavior in animals is important for AI,” he added. “For example, people explore how the neurons in an octopus communicate and coordinate motion to complete certain tasks.” But decoding what’s happening inside the human skull is a tough challenge. Panat pointed out there are about 80 billion neurons in the brain. “Exploring the communication between different parts of the brain at the cellular level is a herculean task,” he said. “At this time, we are limited to the recording of at most 1000 neurons’ talking’ with each other through neural probes. So advances in the recording density and interpretation of the signaling patterns of the neurons has been of immense interest.” As researchers learn more about the similarities between silicon and neurons, computer technology might also one day help treat brain disorders. Maryam Ravan, a professor of electrical and computer engineering at the New York Institute of Technology, recently co-authored studies that used machine learning to improve the treatment of mental health conditions. In one study, the scientists developed a machine learning algorithm to analyze patients’ brain waves and categorize their patterns as biomarkers for bipolar disorder or major depressive disorder. “Increased interest in AI technology has emerged as society has become more accepting and open about mental health,” Ravan said in an email interview. “Given this, I expect that we will continue to see additional studies that leverage forms of AI, including machine learning, to help streamline the treatment and diagnosis of neurological and psychiatric conditions.”
