Israeli researchers create brain-inspired AI algorithm
“Insights of fundamental principles of our brain have to be once again at the center of future artificial intelligence.”
By MAAYAN JAFFE-HOFFMANProcessing an event with multiple objects. (photo credit: Courtesy)
Researchers at Bar-Ilan University have said they have created a new type of ultra-fast artificial intelligence (AI) algorithm based on the dynamics of the human brain.According to a report published Friday in Scientific Reports, the algorithm demonstrates that, despite the human brain computing at a much slower rate than modern computers, it is extremely fast and efficient. As such, the scientists say, “Insights of fundamental principles of our brain have to be once again at the center of future artificial intelligence.”The researchers – led by Prof. Ido Kanter of Bar-Ilan’s Department of Physics and Gonda Multidisciplinary Brain Research Center – used advanced experiments on neuronal cultures and large-scale simulations to demonstrate a type of algorithm that outperforms learning rates achieved to date by state-of-the-art learning algorithms.“The current scientific and technological viewpoint is that neurobiology and machine learning are two distinct disciplines that advanced independently,” he said. “The absence of expectedly reciprocal influence is puzzling.”Kanter explains that the number of neurons in a brain is less than the number of bits in a typical modern computer and that the brain’s computation speed is even slower than the first computers that were invented more than 70 years ago. However, the brain’s learning rules are much more complicated than those of current AI algorithms.“While driving, one observes cars, pedestrian crossings and road signs, and can easily identify their temporal ordering and relative positions,” said Kanter. “Biological hardware [learning rules] is designed to deal with asynchronous inputs and refine their relative information.”This is in contrast to traditional AI algorithms, which are based on synchronous inputs, ignoring the relative timing of different inputs constituting the same frame.Kanter’s conclusion: The disadvantage of the brain’s learning scheme can be an advantage.
