Though he doesn’t believe machine intelligence will exceed human general intelligence anytime soon, artificial intelligence expert Prashant Doshi says that agentic AI systems that can reason and act with minimal human oversight are in the near future.

“There is still some secret sauce missing,” said Doshi, director of UGA’s Institute for Artificial Intelligence (IAI), of the current state of machine intelligence.

Doshi made the comments Tuesday at UGA’s Healthcare 5.0 Conference on precision healthcare through immersive technology and artificial intelligence. 

Large language models, such as ChatGPT, already can write software, for example, and don’t need heavy reliance on (human) software programmers. Beyond those are something called world models, he said. World models are simulations that enable AI to predict and reason about the physical world. Companies are using world models to generate rich, 3D environments for gaming and immersive exploration. Google recently introduced Genie 3, a platform that allows users to generate real time interactivity. Users could, for example, fly over a mountain and could tailor the experience to their liking, such as adding an event or a person into the experience. 

All of these technologies are being closely watched by IAI, Doshi said, as its work is ever-evolving as the field of AI advances. (AI at UGA began unofficially when faculty from Franklin’s linguistics and philosophy departments formed an AI research group in 1984.) IAI, which pulls in 100 faculty from 13 colleges across UGA, offers degrees and certificates in AI and awards seed grants to promising research teams. This year it funded eight seed grants.