While computers have long since surpassed the capabilities of the human brain in certain areas, people have continued to outperform machines at visual object recognition. Now though, computers may be starting to catch up--and the advance has implications for our understanding of the human brain.
A team of researchers at the Massachusetts Institute of Technology has found that the latest artificial deep neural networks (DNN)--computers inspired by biology--are as good at object recognition as the primate brain. If corroborated by further research, the finding has implications beyond the development of new, more effective computers as the DNNs are underpinned by our current understanding of our own brains.
"The fact that the models predict the neural responses and the distances of objects in neural population space shows that these models encapsulate our current best understanding as to what is going on in this previously mysterious portion of the brain," MIT neuroscience professor James DiCarlo said in a statement. The most effective DNNs used layered operations resembling simple and complex cell hierarchy and other designs comparable to the human brain.
Some of the MIT team think the improved understanding of the human brain that arises from work on DNNs could eventually lead to new ways to fix visual dysfunctions in people. The near-term focus of the work is more fundamental, though. The MIT team still need to establish exactly why the performance of similar DNNs varied. Such knowledge would take some of the guesswork out of the development of new systems.