NeuroAI is the name for the research field that sits at the intersection of neuroscience and artificial intelligence. It treats the relationship as a two-way street: neuroscience supplies principles, architectures, and learning rules that can make AI more capable and efficient, while AI supplies models and computational tools that help explain how real brains work. Examples of the exchange include convolutional networks inspired by the visual cortex, reinforcement learning whose central error signal matches dopamine activity, and deep networks used as the best current models of neural responses.
The term was popularized by a 2023 Nature Communications paper led by Anthony Zador and signed by 27 prominent researchers, which argued that fundamental investment in NeuroAI is needed to reach the next generation of artificial intelligence. The authors stressed that current systems lag far behind even simple animals at flexible sensorimotor behavior, and proposed an embodied Turing test, judging artificial agents by how well they act in the physical world rather than by conversation.
NeuroAI does not claim that AI must copy the brain in detail. Its bet is more measured: that the brain is the one existing system that solves general, robust, energy-efficient intelligence, so its organizing principles are a rich source of ideas for engineers, even as those engineers’ models become essential instruments for the neuroscientists.
For a general reader, NeuroAI is the umbrella under which many of the most interesting brain-and-machine stories sit, and it captures a serious bet by leading scientists that biology still has a great deal to teach the builders of AI.