[ad_1]
Going further, the researchers sought to replicate the performance of humans and baboons with artificial intelligence, using neural network models inspired by basic mathematical ideas about what a neuron does and how neurons connect. These models—statistical systems powered by high-dimensional vectors, matrices multiplying layers upon layers of numbers—successfully matched the performance of baboons, but not the performance of humans; they were unable to reproduce the regularity effect. However, when the researchers made a model powered by symbolic elements—the model was given a list of properties of geometric regularity, such as right angles, parallel lines—it closely replicated human performance.
These results pose a challenge for artificial intelligence. Dr. “I love the advancement in artificial intelligence,” Dehaene said. “This is very impressive. However, I believe a profound aspect is missing that is symbol processing” – namely, the ability to manipulate symbols and abstract concepts, as the human brain does. This is the subject of his latest book, “How We Learn: Why the Brain Learns Better Than Any Machine… For Now”
Yoshua Bengio, a computer scientist at the University of Montreal, agreed that current AI lacks anything to do with symbols or abstract reasoning. Dr. Dehaene’s work provides “evidence that the human brain uses capabilities we haven’t yet found in cutting-edge machine learning,” he said.
He said this especially helps us generalize when we combine symbols as we build and reconstruct bits of information. This gap could explain, for example, the limitations of AI, a self-driving car, and the inflexibility of the system when faced with environments or scenarios different from the training repertoire. Dr. Bengio said this is an indication of where AI research needs to go.
Dr. Bengio noted that from the 1950s to the 1980s, symbolic processing strategies dominated “old-fashioned artificial intelligence”. verifying the proof of a theorem). Then came the statistical artificial intelligence and neural network revolution, which began in the 1990s and gained momentum in the 2010s. Dr. Bengio a pioneer Directly inspired by the neuron network of the human brain, this deep learning method
Her Last Research proposes extending the capabilities of neural networks by training them to create or imagine symbols and other representations.
It’s not impossible to do abstract reasoning with neural networks, he said, “we just don’t know how yet.” Dr. Bengio to investigate how human-conscious processing powers can inspire and support the next generation of artificial intelligence. He has a big project with Dehaene (and other neuroscientists). at the end of the day, our understanding of how brains do it,” said Dr. Bengio.
[ad_2]
Source link