A University of Central Florida computer engineering professor is leading research to address the high economic and environmental costs associated with artificial intelligence (AI) foundation models, such as ChatGPT. While these AI models are often free for users, their operation can cost up to $500,000 per run on cloud platforms and generate emissions comparable to burning more than 11,400 pounds of coal.
Jun Wang, a professor at UCF, has received a $600,000 grant from the U.S. National Science Foundation to improve the efficiency and scalability of these models. Wang explained, “Foundation models are transforming industries from healthcare to education to creative works. But the problem is, they are extremely expensive to run. You need a cluster of very powerful graphics processing units (GPUs) and a lot of electricity. That creates [challenges] for small businesses, nonprofits or universities.”
Wang’s research team will focus on three main areas over the next three years: identifying data bottlenecks that waste time and energy; examining scheduling inefficiencies in AI protocols; and working to accelerate start times and communication with GPUs. The goal is to make AI models more effective by coordinating software algorithms with hardware components.
“To reduce the cost and make AI run efficiently and effectively, think of a running AI model like a factory,” Wang said. “You have different teams, machines, workers and logistics, and all of them have to be coordinated. So in AI, those teams are the software algorithm and the hardware components like the GPUs and the operating systems. It just means coordinating across all of them and designing the algorithm and the system together so they work in harmony from the start.”
The research will involve collaboration with the University of Illinois Champaign to build a large GPU cluster and access an AI supercomputer for experiments. Industry partners such as Microsoft, Google, Oracle, and Amazon are also involved in integrating the proposed solutions into their products.
“There’s a lot of exciting things to do in terms of how to make good use of AI, but it’s still expensive for everybody to use and not everybody can afford AI’s computational resources,” Wang said. “By processing the AI models to be faster and more efficient, we reduce the cost and the energy and ultimately allow more people to benefit from the use of AI. And it’s not just about performance and program execution time, it is actually about access.”
Graduate students interested in joining Wang’s research team are encouraged to contact him at jun.wang@ucf.edu.
