Pattern recognition and rules, coupled with business knowledge, help companies identify workflows for new ways to make money. In fact, some analysts have recognized the merits of neurosymbolic AI for some time. While research continues, however, neurosymbolic AI is already delivering value, especially in workflows where structured facts, explainable logic, unstructured data, and natural language need to work together.
The EY-Parthenon neurosymbolic AI team is working with clients to use the EY Growth Platforms’ neurosymbolic platform to plot business scenarios at hyperspeed and find previously unseen paths for growth.
For example, leaders at one global equipment manufacturer were attempting to quantify the impact of new tariffs on their business. Their exercise included an analysis of historical transaction data and market sentiment and an understanding of where the company had price elasticity.
The analysis took the company four months. Using its proprietary data with our neurosymbolic AI platform, which can access over 100 million data sources, the company achieved the same result in one day.
A brief history of tension
“Neuro” (neural networks) and “symbolic” (symbolic reasoning) have battled for attention over the years. Symbolic, with applications like a chess-playing computer that could beat a human world champion, was in vogue from the 1960s through the 1990s. More recently, LLMs have attracted more of the business world’s attention—and investment—due to the prevalence of GenAI.
Today, businesses can turn this systemic conflict between neuro and symbolic into cooperation. Think of a basic online search for a dress to wear at a wedding. Typing in “white wedding dress” uses neural pattern recognition. Typing in “size 10 white wedding dress” adds in the symbolic rule to get a much more useful result.
A side-by-side comparison highlights the advantages of neurosymbolic AI as a growth engine, with symbolic reasoning expanding on the benefits of GenAI by uncovering value in the market that may otherwise be invisible.