NeuroAI research lies at the intersection of neuroscience and artificial intelligence (AI) and is poised to jumpstart a future that includes virtual neuroscience.
Over the past 10 years, the NIH Brain Research Through Advancing Innovative Neurotechnologies® Initiative, or The BRAIN Initiative®, has transformed the field of neuroscience by launching a technology-driven revolution. Yet, we are still barely scratching the surface, and there is a lot more to learn. The BRAIN Initiative recently convened leading experts from neuroscience, artificial intelligence (AI), and related fields at the BRAIN NeuroAI Workshop to explore opportunities enabled by the convergence of brain science and AI.
Over two days, nearly 2,000 attendees from around the world gathered to explore the interface between AI and “natural intelligence,” a manifestation of the architecture of the brain itself.
Four panel discussions illuminated NeuroAI opportunities and challenges that may inform the NIH BRAIN Initiative’s research priorities for the future: Defining NeuroAI for BRAIN; Exploring the Structural and Functional Convergence of Deep Neural Nets and Brains; Advancing Theory for BRAIN through Neuromorphic Computing, Embodiment, and Physical Intelligence; and Toward Reciprocal BRAIN NeuroAI Advances in Intelligent Computing, Robotics, and Neurotechnologies. NeuroAI has the potential to connect across different levels of brain cell types and circuits. This research will help us understand the brain as a biological computer.
A key message from many workshop presenters was the need to study the natural world for clues and inspiration. Natural intelligence that drives animal and human behavior has been shaped by 500 million years of evolution. As noted by workshop participants, even seemingly simple species such as fruit flies express intricate behaviors as they navigate a complex world, highlighting the value of the recently released open access FlyWire connectome, which is already being used in neuromorphic computing studies.
Workshop participants described that an important research goal for NeuroAI is to better understand how the brains of animals (including humans) are so efficient and effective. A brain operates with the power of a low-energy lightbulb to enable speech, thought, emotions, and movement, as well as to continuously sense the external world. In contrast, current AI algorithms can perform more limited tasks and are immensely energy intensive. Forthcoming research that studies brain function and behavior in naturalistic environments will be highly informative. Examples include the BRAIN Initiative’s Brain-Behavior Quantification and Synchronization (BBQS) program.
Neuromorphic computing is a central component of NeuroAI research and was a key workshop focus. It aims to design better computers that can “think” and process information efficiently and quickly, more like the human brain. Neuromorphic chips are specialized hardware systems designed to mimic the structure and functioning of biological brains. They can learn and are energy-efficient because they process millions of signals simultaneously – unlike AI models that work in a sequence and thus gobble enormous amounts of power, which can be expensive. Neuromorphic computers can thus serve as a state-of-the-art model system for answering questions and modeling how the brain processes information and communicates.
While BRAIN Initiative research has always valued interdisciplinary science, NeuroAI research will undoubtedly take this to a new level. Interdisciplinary collaborations across academic labs, federal agencies, and industry partners will be essential for sharing resources, training future NeuroAI researchers, and transforming scientific discovery into health advances and other practical applications.
The NeuroAI Workshop illuminated how much we don’t yet know about how the brain represents the world around us – and how NeuroAI research can be an important avenue on the journey to new knowledge. For example, we do not know or speak the language of the human brain, which, according to many workshop attendees, is largely mathematics. Cracking this source code will be critical for human brain health.
There are many implications for the NIH’s BRAIN Initiative, as noted by the Initiative’s Director, Dr. John Ngai, who provided closing remarks at the workshop. At present, NeuroAI is in its infancy – awaiting new data to simulate, analyze, and pose new research questions about brain function. Widespread availability of open access data across neuroscience will fuel the growth of this exciting new enterprise toward a future biorealistic, predictive model of the human brain.
For more information, see BRAIN NeuroAI Workshop | BRAIN Initiative.