Media Release
From: AAASHow Can Neuroscience Inform AI systems?
Since the early days of artificial intelligence (AI), scientists have turned to neuroscience as a source of guidance. However, even the AI systems of today reflect a highly simplified version of the complex biological neural networks of the human brain, particularly when it comes to human-like learning and perception. In a Perspective, Shimon Ullman discusses the ways in which neuroscience might continue to inform AI technology. One of the biggest challenges in AI development is the ability to achieve human-like learning and perception. However, given our limited understanding of these aspects within our own brains, it is unclear how they could be used to produce artificial human-like cognitive abilities. According to Ullman, this is an area where network systems and the brain differ fundamentally. Human cognitive systems contain innate cognitive structures – evolution-equipped information innate to humans – that help to facilitate the growth of our cognitive skills with very little prior knowledge or experience. Network systems, on the other hand, rely on extended training using large sets of data. However, Ullman suggests that combining artificial deep learning with brain-like innate structures may help guide network models toward more human-like cognitive abilities.