VBFF An Embodied Neuroscience for Cognition and Emotion
Abstract
Recent advances in machine learning have generated great excitement for machine intelligence.Yet, current artificial systems lack key hallmarks of intelligence. In the cognitive domain, unlikebiological systems, they can neither learn new knowledge fast nor can they generalize theirknowledge flexibly. In the emotional domain, they neither experience nor perceive emotions.Currently, the only example of cognitive and emotional intelligence exists within the sphere ofbiology in humans and, to a lesser extent, animals, including our evolutionary relatives, thenonhuman primates. This research proposal aims to use macaque monkeys as a model system toinvestigate the neurocomputational basis of rapid learning, flexible generalization, and embodiedemotional processing, addressing the most salient capability gaps between biological and artificialintelligence.Traditional studies in neuroscience involving monkeys have often used contrived tasks on virtualplatforms. While these tasks have been useful in understanding how individual neurons encodeinformation, especially in sensory areas, they come with drawbacks. Usually, it takes a lot oftraining for the animals to learn these artificial tasks, and once they do, they struggle to apply whatthey ve learned to new contexts. Additionally, the repetitive and monotonous nature of these tasksin the lab does not mimic the real-life situations that trigger strong emotional reactions. Therefore,we need new experimental paradigms to elucidate the neural basis of cognition and emotion.We aim to achieve two main goals. Firstly, we want to explore how the brain quickly learns andapplies knowledge in new situations. Secondly, we re interested in understanding how dynamicbrain-body interactions create emotions. To investigate the first goal, we will innovate a noveltable-top platform that takesadvantage of monkeys natural skills in learning and applying newknowledge when interacting with objects and object affordances. For the second goal, we willdevelop a two-player table-top platform that leverages monkeys# social awareness to provoke awide range ofemotions, enabling us to understand how brain-body interactions mediate andcontrol emotions. To achieve these aims, we ll use cutting-edge technology to monitor brain-wideneural activity, wearable devices to measure body responses, AI-assisted technology to trackanimals# behavior and facial expressions, and advanced analysis methods to elucidate thecomputational principles of rapid learning, flexible generalization, and embodied emotionalprocessing.The expected outcomes of this research include a deeper understanding of the neural foundationsof cognitive learning and emotional processing, providing insights that could enhance artificialsystems and robotics with capabilities more akin to biological intelligence. Furthermore, thefindings could have implications for treating cognitive dysfunction and emotional dysregulationand improving training and operational strategies in contexts requiring cognitive and emotionalresilience. More broadly, a deeper understanding of the computational principles of cognition andemotion in realistic physical and social settings has the potential to revolutionize the science andthe engineering of intelligence.Approved for Public Release.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Feb 24, 2025
- Source ID
- N000142512142
Entities
People
- Mehrdad Jazayeri
Organizations
- Massachusetts Institute of Technology
- Office of Naval Research
- United States Navy