Drew Zhang Did AI Before AI Was Cool

The Alfred J. Verrecchia Endowed Chair in Artificial Intelligence and Business Analytics in the College of Business, Drew Zhang sees artificial intelligence not as a threat, but as a tool that today’s business students must learn to use.

It was late autumn in 2022, and Zhu “Drew” Zhang was busy tying up loose ends at Iowa State University, where he had spent the last eight-and-a-half years on the faculty. He was starting a new position as a professor at URI in January and had hoped to enjoy a little rest before arriving in Kingston, but he soon found his email and voicemail overflowing with messages from his new colleagues at URI’s College of Business, many of whom he had yet to meet in person. They were reaching out to Zhang to ask about ChatGPT, a then-new app—released that October by OpenAI—that allows users to chat with an artificial intelligence.

“Before I had even officially started my job at URI, I was getting all these emails and calls from faculty who knew I worked with this kind of stuff and wanted me to tell them about it,” Zhang says.

ChatGPT wasn’t the first AI chatbot by a longshot, but compared to Siri, Alexa, and other familiar AI systems, its capabilities seemed downright magical. As a large language model—a type of AI that is trained on astronomical amounts of textual data—ChatGPT could write short stories, summarize scientific papers, solve calculus problems, and write code in a way that seemed entirely human.

For Zhang, who has spent the past 25 years working on text-based AI—a subspecialty known as natural language processing—the arrival of large language models wasn’t a surprise. In fact, he and many of his colleagues in the field had done the foundational work that made such models possible. But what was surprising was how quickly the new tool graduated from the stuff of dry academic conferences to a global phenomenon. “When I was a doctoral student, people weren’t really comfortable talking about AI, but now everyone is talking about it as a result of ChatGPT,” says Zhang. “I consider myself lucky that I started working on this before it was cool.”

Within weeks of its release, ChatGPT claimed the mantle of the fastest-growing consumer app ever (besting TikTok, Instagram, and other blockbuster social media platforms), and only two years after its release, it’s used by more than 200 million people every week. Many of Zhang’s new URI colleagues saw that ChatGPT was poised to make a huge impact on the business and financial markets, but it wasn’t at all obvious just what this impact would be. It was a familiar perspective to Zhang, who, over the course of his career, has grown used to being something of a black sheep in the business world.

At a high level, Zhang specializes in extracting business insights from massive amounts of data—far too much data for humans to comprehend. He does this using machine learning models—AI models that “learn” by looking for patterns of interest in training data and then applying the resulting statistical model to find similar patterns in real-world data.

Business leaders are used to working with large datasets. It’s standard fare on Wall Street and any mid- to large-sized company. The difference, however, is that these businesses traditionally work with large, structured datasets—think commodity prices or inventory lists—that are in nice tidy rows that make them comparatively easy to analyze and extract valuable information from.

The data that Zhang works with, however, is unstructured language data pulled from sources like social media or Amazon product reviews. And language data—especially the type of informal language found on social platforms—is messy. Zhang and his collaborators must contend with grammatical errors, formatting variation, and missing context that is critical for understanding whether, say, a product review is ironic or sincere. “Language data used to be considered a second-class citizen in the business world,” says Zhang. “Nobody knew how to work with it, and the value wasn’t immediately obvious.”

Large language models like ChatGPT changed everything. For most of his career, Zhang often had to build his own AI models for text analysis, which was laborious. Now, with a consumer tool like ChatGPT, anyone can extract meaningful insights from large amounts of text. And if you’re an expert like Zhang who knows how to work with large language models at a technical level, it’s a bit like being granted a superpower.

Lately, Zhang has been focused on applying his AI research to marketing and finance, two areas that have a lot to gain from scalable text analysis. As an example on the marketing side, Zhang says businesses and their customers stand to benefit from the ability to innovate on product design based on user reviews mined from the internet. Beyond just being able to tell whether customers like a product, businesses can do a more detailed analysis to understand what, exactly, customers like or dislike. On the financial side, Zhang is exploring how language data can be mined and used to predict market volatility, which can help professional investors and everyday people better manage risk—which might be particularly useful for people nearing retirement age who are counting on making their savings last.

Zhang’s work on AI-driven business analytics caught the attention of Alfred J. Verrecchia ’67, M.B.A. ’72, Hon. ’04, board chairman and former president and CEO of Hasbro Inc., who knew firsthand the importance of big data in business and the challenges of making it actionable. A gift to URI from Verrecchia and his wife, Geraldine Verrecchia, created an endowed chair for artificial intelligence and business analytics. There were a lot of great candidates, but Zhang stood out because of the increasing practical relevance of his work for URI students heading out to work in the business world.

“My passion is to make sure students are prepared and have the skills necessary to compete in today’s environment, and artificial intelligence and business analytics are an important part of that,” says Verrecchia. “There are people who can do great research, and I have a lot of respect for those people, but you also need to teach students, which means relating to them and making sure they have a good experience. After meeting Drew, I felt very strongly that he would be terrific at that.”

Zhang, like Verrecchia, believes the new wave of AI will have a profound impact on the way we do business, and on the world at large. But he’s also been around long enough to have seen AI hype cycles wax and wane. Zhang isn’t worried about AI taking over the world—or even most people’s jobs, for that matter—anytime soon. He says the important thing is to help students prepare for real-world applications of AI technologies. In opposition to oft-cited fears about ChatGPT and the use of similar tools in higher education, Zhang actively encourages his students to experiment with the tool and his colleagues to rethink their pedagogy.

“AI isn’t ready to replace human intelligence yet,” says Zhang. “Math teachers were scared of students using calculators without having to learn the math systematically, but math education has only been empowered by the invention of calculators,” says Zhang. “I think of these AI models in the same way—it’s a new family of tools, and we have to educate students to use them.”

—Daniel Oberhaus

PHOTO: BEAU JONES