Deep Neural Networks: A New Framework for Modeling Biological Vision and Brain Information Processing

What is it about?

The brain-inspired technology of deep neural networks has recently enabled computers to recognize objects in images, markedly reducing the gap between human and machine recognition abilities. Deep neural networks have also been successful at speech recognition, language translation, speech synthesis, and a range of other artificial intelligence applications. This article argues for the use of this technology to model of how biological brains perform complex feats of intelligence.

Featured Image

Why is it important?

Computational neuroscience has been very successful at explaining low-level sensory and motor representations as well as some of the building blocks of higher-level processes, such as decisions. However, at this level of rigor, we have not been able to explain complex cognitive processes that require rich knowledge about the world. Like artificial intelligence before it, computational neuroscience must embrace the fact that ample world knowledge is essential to cognition. This means that machine learning, which provides the knowledge needed for intelligence, is central to the computational neuroscience of complex cognition. Neural network models, although they abstract from many biological details, provide a key link that will enable us to explain how complex cognition works in biological brains, including the human brain.