Temple architecture

Born in the 80s, Temple’s NARS project is still a leader in general artificial intelligence

Did you know that Temple University is home to a decades-old AI systems project?

The NARS project (Non-axiomatic reasoning system) is an artificial general intelligence (AGI) system focused on reasoning, learning, planning, perception and temporal/procedural control. The project was founded in the 1980s by Dr. Pei Wangwho is currently an associate professor and researcher at North Philadelphia University.

Peter Isaev and Patrick Hammer worked on the implementation and practical application of the NARS project as research assistants within the College of Science and Technology Computer and Information Science Department. Hammer earned his Ph.D. in Computing and Information Science from Temple in 2021; Isaev is pursuing his now.

hammer says Technically the project is unique among others in artificial intelligence because it learns from experience in real time. When the system works, it can test assumptions to see what works and what doesn’t, making project implementation practical.

“It’s really about real-time uncertainty, reasoning and learning,” Isaev added, “which stands out from the rest of AGI systems because real-time and uncertainty are the key aspects of NARS.”

Many successful machine learning systems are special-purpose systems for performing specific tasks, Hammer said, but AGI is about building reasoning systems that learn from experience and perform many tasks. The project is currently working on integrating NARS into robotics.

Future AI logo. (Courtesy image)

NARS recently received a grant from a DC-based AI company The AI ​​of the future, launched in January of this year. Isaev said members of the NARS project met Future AI at an AGI conference and are now acting as consultants for Future AI on its technology. Funding details were not disclosed.

“By using the general-purpose reasoning model developed for the NARS project, we believe we can strengthen our efforts to ensure that Future AI’s graph-based AI system will exhibit desirable cognitive capabilities,” said the founder. and CEO of Future AI. Charles Simon said in a statement.

Future AI’s technology is very different from NARS, but the AGI community is very small, so community members are trying to find common ground to continue their research, Isaev said. The purpose of the collaboration is to “provide knowledge on NARS and advise on how to integrate it into their current architecture”.

The NARS project received the grant in early fall and the collaboration will continue until the end of this semester in December. Temple researchers said the collaboration was going well and both groups were able to see and address the challenges they faced.

However, Future AI is not the only company the NARS project has partnered with: the project has worked with companies like Nasa and Cisco as well. Overall, Isaev said, AI is currently limited and requires time, commitment and partnerships to achieve goals within AGI.

Sarah Huffman is a 2022-2023 corps member of Report for America, an initiative of The Groundtruth Project that pairs young journalists with local newsrooms. This position is supported by the Lenfest Institute for Journalism. -30-