Home » Technical Topics » Knowledge Engineering

Scaling knowledge graphs and neurosymbolic AI

  • Alan Morrison 

An interview with Amit Sheth, founding director, Artificial Intelligence Institute, University of South Carolina

Amit Sheth

Amit Sheth has survived and thrived through multiple AI winters because of a broad and well-informed perspective. He understands how AI needs to evolve for the sake of accuracy and trusted human-machine collaboration. Knowledge representation — the symbolic AI technology behind knowledge graphs — has been one of his main specialties for many years. 

After joining Bell Labs in 1989, Amit led the implementation of a faceted search engine called Adapt/X Harness for Bellcore (later Telcordia, now a unit of Ericsson called iconnectiv) via a Mozilla browser back in 1993, before e-commerce and most of the commercial web had materialized. In 2000 at Taalee (later merged with Voquette), he led the development of MediaAnywhere, a pioneer in audio and video web semantic search. He founded Taalee and has founded many other startups since then.

In 2016, while director of the Kno.e.sis Center at Wright State University, Amit co-founded Cognovi Labs, an AI company that leveraged a natural language model called Twitris developed under Federal research grants at Wright State. Cognovi Labs uses psychology and technology methods together to measure emotional and decision-making factors.

Currently, Amit’s focus is neurosymbolic AI, the effective blending of knowledge graph and probabilistic neural network (such as deep learning and generative AI) methods. He founded the Artificial Intelligence Institute at the University of South Carolina in 2019, where he continues to serve as director. As a longtime director of academic AI R&D institutes, he is proudest of the dozens of PhD candidates he has overseen over the decades. 

“Today,” Amit says in the edited interview linked below, “we have a platform called EMPWR (pronounced ‘empower’) to help build knowledge graphs from unstructured, semi-structured and structured knowledge sources with a lot of automation. And so one of the challenges is building a knowledge graph of 10 million nodes and many, many relationships without the kind of tooling EMPWR provides.”

At the AI Institute of the University of South Carolina, he says, “We are serving the needs that were not very well satisfied by data driven-alone approaches. There is increasing growth of understanding by not only academia but also industry that knowledge is playing an important, necessary role.”

Hope you enjoy the podcast.

DSC podcast interview with Amit Sheth

Leave a Reply

Your email address will not be published. Required fields are marked *