Examining the bridge between neuroscience and AI

A study analyzing the connection between neuroscience and artificial intelligence reveals that the link is real and growing. However, the relationship is narrow and heavily reliant on tools. The two fields have a long history, with early AI attempts drawing inspiration from the brain. Recent advancements in AI, while not directly stemming from the brain, still benefit from concepts that were absorbed decades ago.

Despite this, neuroscience remains the only known system of general intelligence with efficiency and flexibility that AI has yet to achieve. There is a growing consensus that the two fields should collaborate, with AI offering rigorous foundations and tools to neuroscientists. This collaboration is evident in dedicated institutions, endowments, and events.

However, the actual impact of this merger is difficult to measure. The research analyzed over 458 million publications and the 2.9 billion links between them, tracing the flow of knowledge between the fields. The findings show that the connection is real and expanding, but it is still limited to specific groups and institutions.

Most of the knowledge that moves from AI to neuroscience are tools, while much of the knowledge that moves from neuroscience to AI are classical concepts. The drive behind this influence comes from a small, fixed set of classic neuroscience papers that provide important conceptual foundations for artificial neural networks and deep learning.

The recent increase in AI citing neuroscience papers is due to the widespread adoption of AI tools in analyzing neural data. However, this may also be partly due to neuroscience's shrinking share in a booming AI literature. The future of this collaboration remains to be seen, but the scientific community must turn these early signs of growth into sustained progress.