AI’s growing presence in scientific research is reshaping how complex questions are tackled, particularly in fields like neuroscience. According to a new article by Kenneth Harris, a professor of quantitative neuroscience at University College London, the key to leveraging AI’s potential lies in collaboration—turning hard scientific questions into solvable, checkable proposals that AI can iterate upon.
AI’s strength is its ability to perform trial-and-error efficiently, systematically testing various solutions. This method is particularly useful for neuroscience, where research often involves high-dimensional data and complex patterns. Rather than relying solely on human intuition, AI systems can rapidly generate multiple hypotheses and test them in real-time, producing results at a speed far beyond traditional methods.
For example, Harris’s lab used AI to solve a challenging issue in visual neuroscience. By turning the problem into a “game” for the AI, they were able to use a fixed scoring loop that evaluated AI-generated solutions based on their accuracy in predicting neuronal responses to visual stimuli. After hundreds of iterations, the AI system arrived at a new model, the “stretched exponential,” which explained the high-dimensional population code in a way previous models could not.
AI’s value in such contexts is not about creating breakthroughs humans could never imagine. Instead, it excels at taking established knowledge, applying it across a range of scenarios, and finding incremental improvements faster than humans alone could. This collaboration with AI allows researchers to build on insights that may have taken years to uncover through traditional methods.
However, while AI’s contributions to neuroscience are invaluable, its current role remains a supportive one. Human researchers still drive the questions, curate the data, and interpret the results. AI provides the tools to accelerate discovery but does not replace the need for human creativity, insight, and scientific rigor.
As Harris suggests, the future of neuroscience will increasingly involve AI as a collaborator—acting as a constant, tireless worker in the background, refining hypotheses and proposing new models based on data. This shift could revolutionize not just neuroscience but many scientific fields, enabling faster, more robust discoveries that build on human expertise.
