August 8, 2022

SEC AI Collaboration Advances Computing Benchmarks for Researchers

By: SEC Staff
SECU (Twitter: @TheSECU)

BIRMINGHAM, Ala. — Artificial intelligence is everywhere.

As AI and machine learning increase in daily life activities and in the roles they play in the workplace, the Southeastern Conference and its 14 member universities are committed to preparing graduates for career opportunities while enhancing faculty and staff involvement in those fields.

The SEC Artificial Intelligence and Data Science Consortium was created in 2021 to move these initiatives forward. One way the SEC universities are working together is by sharing resources such as the University of Florida’s HiPerGator Artificial Intelligence Supercomputer.

Texas A&M University’s Dhruva Chakravorty and Lisa M. Perez are part of a team led by Honggao Liu, executive director of the High Performance Research Computing (HPRC) Facility, that used the system to build benchmarks for AI researchers. Supported by the university’s Accelerating Computing for Emerging Sciences (ACES) project through the National Science Foundation, the group ran the same code on supercomputers at UF, Texas A&M and other institutions to establish baselines.

Remote access to the HiPerGator and other machines allowed the team to compare differences in machine architecture, the performance of Graphical Processing Units and Intelligence Processing Units, and to evaluate what went well and what did not on the computers. What the researchers discovered was outlined in a recently published academic paper.

The paper was tremendously collaborative, as we shared the results produced on machines at four different sites,” said Chakravorty, director for user services and research at HPRC. “These collaborations are very important as we move ahead because machines are getting very complicated. We are trying to simplify these different and diverse resources for researchers, so this helps them focus on the science behind their research and not the computing.”

In addition to the publication, the outcomes from the team’s research were also presented at the Practice and Experience in Advanced Research Computing Conference in Boston.

“The HiPerGator machine really helps with how we use AI to develop new technologies and accelerate scientific research,” Chakravorty said. “It’s a critical piece in this educational triangle of how we build, apply and use AI technology. Having this technology in our Conference can inspire scientists to solve previously un-addressable challenges. That is just incredible.”

Perez, associate director of advanced computing enablement at HPRC, said educational outreach is one of many positives of the SEC consortium since bringing AI forward and sharing ideas will help all universities involved. In addition, having access to these shared instruments and ideas will help build AI knowledge throughout the Conference and beyond.

“We are developing these relationships and bringing schools together around how we share AI resources and bring that curriculum to all levels,” Perez added. “We are discovering how we bring the fundaments and training of AI/machine learning into every college because it can be really important in advancing every discipline.”

Believed to be the first athletics conference collaboration to have such a focus, the SEC Artificial Intelligence Consortium enables SEC universities to share educational resources, such as curricular materials, certificate and degree program structures, and online presentations of seminars and courses; promote faculty, staff, and student workshops and academic conferences – including a recent event at Auburn University; and seek joint partnerships with industry.