Tampa Bay Times: UF and other research universities will fuel AI. Here’s why | Column

In the global AI race between small and major competitors, established companies versus new players, and ubiquitous versus niche uses, the next giant leap isn’t about faster chips or improved algorithms. Where AI agents have already vacuumed up so much of the information on the internet, the next great uncertainty is where they’ll find the next trove of big data.

The answer is not in Silicon Valley. It’s all across the nation at our major research universities, which are key to maintaining global competitiveness against China.

To teach an AI system to “think” requires it to draw on massive amounts of data to build models. At a recent conference, Ilya Sutskever, the former chief scientist at OpenAI — the creator of ChatGPT — called data the “fossil fuel of AI.” Just as we will use up fossil fuels because they are not renewable, he said we are running out of new data to mine to keep fueling the gains in AI.

However, so much of this thinking assumes AI was created by private Silicon Valley start-ups and the like. AI’s history is actually deeply rooted in U.S. universities dating back to the 1940s, when early research laid the groundwork for the algorithms and tools used today. While the computing power to use those tools was created only recently, the foundation was laid after World War II, not in the private sector but at our universities.

Contrary to a “fossil fuel problem,” I believe AI has its own renewable fuel source: the data and expertise generated from our comprehensive public academic institutions. In fact, at the major AI conferences driving the field, most papers come from academic institutions. Our AI systems learn about our world only from the data we offer them.

Current AI models like ChatGPT are scraping information from some academic journal articles in open-access repositories, but there are enormous troves of untapped academic data that could be used to make all these models more meaningful. A way past data scarcity is to develop new AI methods that leverage all of our knowledge in all of its forms. Our research institutions have the varied expertise in all aspects of our society to do this.

Here’s just one example: We are creating the next generation of “digital twin” technology. Digital twins are virtual recreations of places or systems in our world. Using AI, we can develop digital twins that gather all of our data and knowledge about a system — whether a city, a community or even a person — in one place and allow users to ask “what if” questions.

The University of Florida, for example, is building a digital twin for the city of Jacksonville, which contains the profile of each building, elevation data throughout the city and even septic tank locations. The twin also embeds detailed state-of-the-art waterflow models. In that virtual world, we can test all sorts of ideas for improving Jacksonville’s hurricane evacuation planning and water quality before implementing them in the actual city.

As we continue to layer more data into the twin — real-time traffic information, scans of road conditions and more — our ability to deploy city resources will be more informed and driven by real-time actionable data and modeling. Using an AI system backed by this digital twin, city leaders could ask, “How would a new road in downtown Jacksonville impact evacuation times? How would the added road modify water runoff?” and so on.

The possibilities for this emerging area of AI are endless. We could create digital twins of humans to layer human biology knowledge with personalized medical histories and imaging scans to understand how individuals may respond to particular treatments.

Universities are also acquiring increasingly powerful supercomputers that are supercharging their innovations, such as the University of Florida’s HiPerGator, recently acquired from NVIDIA, which is being used for problems across all disciplines. Oregon State University and the University of Missouri, for example, are using their own access to supercomputers to advance marine science discoveries and improve elder care.

In short, to see the next big leap in AI, don’t immediately look to Silicon Valley. Start scanning the horizon for those research universities that have the computing horsepower and the unique ability to continually renew the data and knowledge that will supercharge the next big thing in AI.

Alina Zare, Ph.D., teaches and conducts research in machine learning and AI as a professor in the Department of Electrical and Computer Engineering at the University of Florida. She also serves as the director of UF’s Artificial Intelligence and Informatics Research Institute.

This article was originally published by the Tampa Bay Times.