Brown University researchers are starting a new institute to develop AI assistants capable of having “trustworthy, sensitive, and context-aware interactions” with humans, aimed at improving their mental health. The AI research institute, announced Tuesday, is among five universities that were awarded grants totaling $100 million from the US National Science Foundation, in partnership with Intel and Capital One, to boost US competitiveness and align with the White House’s “AI Action Plan.”
Brown’s institute is developing a new generation of chatbots that can better interpret an individual’s behavioral needs and respond in real time in ways that create a safe environment.
“Any AI system that interacts with people, especially who may be in states of distress or other vulnerable situations, needs a strong understanding of the human it’s interacting with, along with a deep causal understanding of the world and how the system’s own behavior affects that world,” Ellie Pavlick, Brown’s associate professor of computer science who’s leading the project, said in the release.
AI therapy has grown in popularity partly because it’s low cost and more accessible to the general public, but Stanford University research has also warned that current large language models introduce bias and failures that could have “dangerous consequences.” Brown’s institute aims to develop a safer AI system with analogous capabilities based more on cognitive science and neuroscience compared to the existing LLMs that generate text responses based on prior words or user inputs.