Cornell University researchers from the Department of Psychology's Computational Physiology Lab and the Cornell University AI for Science Institute have developed a novel AI strategy inspired by the human brain's olfactory system to efficiently process noisy sensory data. The brain's ability to sift through complex sensory information with minimal energy consumption offers valuable insights into designing advanced AI systems. The human brain excels at organizing chaotic input from the environment, transforming it into meaningful and reliable representations. For example, it can easily distinguish the aroma of tomato sauce amidst the diverse scents in a bustling restaurant. This proficiency is rooted in the brain's olfactory system, which involves the olfactory epithelium, the olfactory bulb, and downstream brain areas. The olfactory epithelium senses chemicals, while the olfactory bulb sorts and refines this raw data before passing it to other brain regions for further interpretation. The researchers focused on the initial computations performed by the olfactory epithelium and the outer layer of the olfactory bulb. These layers act as a "firewall," organizing and constraining sensory input to ensure it is manageable by deeper brain structures without loss of crucial information. This is essential for the deep layer of the olfactory bulb, which is highly dynamic and specialized for learning about different odors. However, it requires well-structured inputs to function effectively. The team's findings extend beyond olfaction. They propose a generic regularization mechanism applicable to various types of sensory data, making the approach broadly relevant for AI and robotics. The researchers discovered that the brain uses spike-phase coding, a method where neurons communicate by precisely timing their signals. This energy-efficient strategy also supports stable learning and handles noisy and sparse data, suggesting parallels with modern techniques like quantization-aware training in machine learning. Postdoctoral researcher Roy Moyal, the first and co-corresponding author of the study, aims to develop brain-inspired neural networks that surpass current AI systems in learning capabilities, energy efficiency, and size. He envisions lightweight, autonomous AI agents capable of detecting hazardous materials, adapting to local conditions, and processing data quickly without relying on external networks. Such systems could operate on small, purpose-built devices, enhancing their practical utility. The study, published in Scientific Reports, highlights the potential for neuromorphic design in creating AI systems that match the brain's performance. According to Moyal, the algorithm derived from these principles can regularize any data its sensors can encode, and prior knowledge can enhance its performance. Key contributors to the research include Kyrus Mama '21 and Matthew Einhorn from the Computational Physiology Lab, and Ayon Borthakur from the Indian Institute of Technology, Guwahati. Industry insiders view this research as a significant step toward more efficient and adaptable AI systems. The development of brain-inspired algorithms could revolutionize how robots and other devices process sensory information, leading to more robust and energy-efficient solutions. Thomas Cleland, a professor of psychology, emphasized that while there is much work ahead, this study lays a foundational understanding that could catalyze future advancements in neuromorphic engineering and AI. The Computational Physiology Lab at Cornell University is known for its interdisciplinary approach, combining insights from neuroscience, psychology, and engineering to solve complex problems in AI. The Cornell University AI for Science Institute supports cutting-edge research aimed at integrating AI with scientific disciplines, fostering innovation and collaboration. Together, these institutions are pushing the boundaries of what AI can achieve, particularly in areas requiring real-time data processing and low energy consumption.