San Antonio Express-News (Sunday)
Working to make AI a reality
UTSA researcher seeks to draw San Antonio’s underrepresented students into the field
In the farming village where Dhireesha Kudithipudi was born in southern India, most of the girls were married by their teenage years.
But Kudithipudi’s mother, who had a high school diploma, encouraged her daughter’s academic studies. And her father, an electrical foreman, introduced her to the technology he encountered in his travels.
The homegrown motivation helped her decide to pursue the degrees in computer and electrical engineering that led her to San Antonio, where she leads a group of researchers attempting to build artificial intelligence — computer systems that can adapt and learn how to perform tasks like humans.
“We see AI as technology that is going to help humans,” she says. “Whether we accept it or not, these AI systems are everywhere around us, when you’re doing your Netflix search or buying something on Amazon or Target. AI is around us already in ways that we don’t see or realize.”
From her post as director of the Matrix: AI Consortium for Human Well-Being at the University of Texas at San Antonio, the 43-year-old has another goal: recruiting diverse talent from San Antonio.
“It hits home for me, since
I’m first-generation,” she said. “It’s our responsibility to train these students and equip them with skills.”
Since its inception two years ago, Matrix has pulled together researchers from various academic disciplines from UTSA, UT Health San Antonio, Southwest Research Institute and Texas Biomedical Research Institute. Their mission: “To conduct transformative research in the design, use, and deployment of AI that enhances human life, and to offer rigorous research training opportunities that transcend disciplinary boundaries,” according to its 2021 annual report.
Kudithipudi leads the charge on classified and nonclassified research for clients including the U.S. Air Force and Sandia National Labs in New Mexico, and Xilinx, a tech and semiconductor manufacturing company in California that was just acquired by AMD.
“This is a very unique time that we are in right now with AI technology because we are working with industry,” she said. “We are helping them solve these problems, and the students are able to be on the front end.”
The Matrix consortium
The Matrix consortium is made up of researchers and professors from a wide range of academic fields. They include medical engineering, neurology, computer science, mathematics, infectious disease, geological sciences, urban planning, modern languages, radiology, automation and psychology.
“We are trying to bring researchers, scientists and students together under one umbrella to design and deploy AI solutions to help human wellbeing in any form that we can help boost human performance, help humans perform tasks more efficiently,” she said.
And Kudithipudi is focused on recruiting diverse talent from San Antonio.
She noted that just 5 percent of the current AI workforce in the U.S. is from underrepresented groups. That means Matrix has an opportunity to reach young people on a campus where more than 65 percent of the student population comes from such groups.
Her outreach has found many avenues. Kudithipudi has been an adviser for Project Lovelace, a program offering opportunities for young women in STEM fields to engage in research and participate in seminars and workshops. Under her leadership, Matrix also hosts educational programs for students and the public. Every fall and spring semester, AI researchers give seminars on algorithms, theory, systems and autonomy.
Matrix has also partnered with MISI Dreamport and BigBear.ai, both from Maryland, to run a symposium on AI and quantum computing to discuss how the technologies will fundamentally change how people interact with data.
Matrix next is partnering with the city of San Antonio Research & Development League to collaborate on innovation-related projects for the community.
Rearing a researcher
Kudithipudi recalled her mom “had big dreams” for her as she was growing up in southern India,
And her father — her “role model” — bought Kudithipudi her first computer when she was in the sixth grade, a rarity in the farming village. He also gave her and her younger brother Atari-branded video games. They were told to figure out how to play the games without instruction manuals.
“Our father gave us the opportunities to look beyond where we were,” she said.
When she finished high school, the family moved to the city of Hyderabad — now known as a tech capital in south-central India.
Following her father’s professional path, she received her bachelor of technology degree in electrical and electronics engineering from Acharya Nagarjuna University in 1998.
“It was hard-core designing and thinking about power grids and hands-on technology,” she said. She wasn’t thinking about AI at the time.
Between 1999 and 2002, she pursued her master’s degree in computer engineering at Wright State University in Ohio. While her interest was in building hardware for computer chips, she took a course in neural networks, which involved studying computer systems that reflect the behavior of the human brain to recognize patterns in data sets.
In 2002, she was accepted to a post-doctorate program for electrical and computer engineering at UTSA. “I was working still on hardware design but trying to make systems lowpower and energy efficient,” she said.
But in her second year of the PhD program, she read a paper from MIT professor Ron Weiss, a pioneer in synthetic biology. Its work, using DNA synthesis and gene sequencing, is making cells to increase food production, combat disease and generate energy, as some examples.
“He showed how computing happens in biology,” she said, adding that she went to local biology professors to work on such a project. She was willing to extend her studies to do so, but found no one was working on the subject.
Kudithipudi next worked as a professor from 2006 to 2019 at the Rochester Institute of Technology in New York. She was still interested in designing energy-efficient computing but shifted into biology, again. This time, she was looking at how brains process information and thinking about how she could bring that processing into computing. As director of the Neuromorphic AI (Nu.AI) Lab there, her research team created
AI platforms inspired by the brain that had continual learning abilities.
For the fall of 2019, UTSA recruited seven researchers as part of a “cluster hiring initiative in AI,” according to the university. That year, Kudithipudi was brought onto campus to lead Matrix and serve as the Robert F. McDermott Chair in Engineering and professor of electrical and computer engineering.
Today, Kudithipudi lives with her husband, Surendar, an engineer she calls “her rock.” Together, they are raising two boys, ages 12 and 9.
Typically, she does not bring her work home. But she laughed talking about how she brought “circuits” to her younger son for his birthday. “I made them build circuits with PlayDoh and the kids had a blast,” she said, noting that she was introducing her children to forms of tech, just as her father did for her.
In the lab
On a recent afternoon, Kudithipudi sat in her Neuromorphic AI Lab — one of about 25 labs tied to Matrix on the UTSA campus.
Its name explains its function: “Neuromorphic” refers to a method of computer engineering focused on designing hardware and software modeled on the human brain and nervous systems.
“The lab is interested in designing the next generation of AI systems that can solve complicated and natural tasks similar to what humans do,” she said. “We want to do that in the most energy efficient way.”
As Kudithipudi explains it, computers are better at number-crunching while humans excel at recognizing patterns.
Her team “is trying to give the ability to machines that do not have the natural strength of pattern recognition, organic problem-solving and anomaly detection,” she said. And they’re trying to figure out how to make computers learn over the course of many years, as humans do. “Lifelong learning is a grand challenge of AI,” she added.
Nearby, lab researchers work with an AI model called “Neurogenesis” that’s learning continuously without forgetting, a contracted project for the Department of Defense.
They design 3D computer simulations, then use code to connect the model to a so-called agent: a digital drone, robot on
Mars or something anthropomorphic like a spider. Then, they gave the agent a task and watch how it interacts with the environment.
Kudithipudi oversaw researchers who programmed a digital spider to perform several tasks. In one, a big spider was trained to chase a small spider around a maze. After about 50 tries, the smaller spider learned how to get away from its predator.
In such projects, researchers input data and tweak algorithms to modify the AI’s behavior to perform the task. They say it’s like evolution.
“When we are babies, we are learning the language — the names of things — and what is dangerous and what is OK to touch. A lot of these things we learn in our early stages of life and then we use that knowledge for the rest of our lives,” Kudithipudi said. “It’s kind of similar with AI.”
‘That’s the future’
Researchers make the models nonapplication specific. The goal is for the AI system to learn a sequential task, for example. Then, a client like the Department of Defense can use the system for any number of tasks.
“Whether it’s in a battlefield, or some other applications we cannot talk about,” Kudithipudi said.
The researchers are trying to improve the speed at which an AI system learns to cut down on power generation and ultimately cost for real-world applications.
“The hardware is a computer chip that can go into autonomous vehicles, drones, sensors or satellites in space,” Kudithipudi said.
She called “dire” the need for it to operate with low power.
“You would want more and features on your cellphone in the coming years,” Kudithipudi said. “Currently, a lot of AI models on your cellphones, like facial recognition . ... That’s the future we are working toward: to bring this technology in compact form to the users.”
Such challenges drive her professionally.
“Early on in my career, I was working on the research problems,” she said. “But I’ve realized what makes me feel content or happy to go to work is to see the solutions that we are building are making a difference in the world.”