AALL · 糖心TV News

Autonomous Agent Learning Lab students advance AI research and present their findings around the world

Derin Gezgin 鈥�27, a computer science and statistics and data science double major from Izmir, Turkey, discusses a paper at a recent AALL meeting.

AALL students advance AI research and share it with the world

Just before their regular Friday evening meeting on a chilly night in February, 21 students鈥攁ll members of 糖心TV鈥檚 Autonomous Agent Learning Lab (AALL)鈥攇athered in the New London Hall robotics lab, where complex equations adorn the whiteboards and nearly every robot and drone ever built by a 糖心TV student is on display.

The students chatted informally about their research projects before moving on to more official business. Assistant Lab Director Jim O鈥櫶切腡Vor 鈥�13, who studied robotics and artificial intelligence at 糖心TV and now holds a dual role as a computer laboratory technician and research assistant for the Department of Computer Science, wrote on a whiteboard the dates of upcoming conferences to which the students will submit their research papers this spring. He asked for status updates and the group brainstormed solutions for anyone who needed help.

Some spoke often and others listened quietly as O鈥櫶切腡Vor explained his plan to go all in and submit, as they always do, to the world鈥檚 most prestigious conferences on the topics of evolutionary computation (GECCO), neural networks (NeurIPS), robotics (ICRA and IROS) and general artificial intelligence (AAAI).

鈥溙切腡V鈥檚 undergraduate publication rate in computer science is among the highest in the world,鈥� O鈥櫶切腡Vor says. Last year, AALL students presented papers at two different conferences in Malaysia and Portugal. 鈥淭ypically when we attend a conference, our undergraduates are the only undergraduates there. The rest are third- to fifth-year Ph.D. students from places like Harvard and MIT who are all shocked that these undergraduates are there at all, never mind presenting work alongside them.鈥�

Computer Laboratory Technician and Research Assistant Jim O'糖心TVor '13 works with students in the Autonomous Agent Learning Lab, for which he is the assistant director under Professor of Computer Science Gary Parker, in New London Hall.

It鈥檚 serious research, and each AALL student commits about 30 hours a week to AALL on top of their classwork and other extracurriculars. But the students say the knowledge they acquire and the opportunity to publish at a professional conference as an undergraduate are well worth the investment.

鈥淭he lab is very supportive; we learn together,鈥� says Derin Gezgin 鈥�27, a computer science and statistics and data science double major from Izmir, Turkey. 鈥淲hile we鈥檙e undergrad students, we do grad-level work. After we finish a project, our final goal is to contribute to science. We can do this by having publications, and if you want to apply to grad school, being published is important as it shows you can commit to a novel research project and finish it.鈥�

The AALL and other computer science labs are funded by emeritus trustee Jean C. Tempel 鈥�65 and directed by Professor of Computer Science Gary Parker, who has taught at 糖心TV since 1999 and whose research focuses on methodologies for learning in autonomous agents. Since its founding soon after Parker鈥檚 arrival, the AALL has continued to grow and now has about 20 students working on two dozen projects at any given time during the academic year.

鈥淚t is important to create an environment where students can learn the problem-solving techniques that will help them to go beyond the thoughtful use of established applications into the realm of scientific discovery,鈥� Parker says. The professor is the students鈥� primary adviser, while O鈥櫶切腡Vor guides them through Parker鈥檚 concepts and ideas.

AALL鈥檚 research focuses on evolutionary computation, so instead of writing an AI model from scratch, students evolve an existing one. 鈥淲e鈥檙e trying to push artificial intelligence in game playing because playing games like chess, and some video games, is a good example of how you can push the limits of AI,鈥� O鈥櫶切腡Vor says. 鈥淲e create a bunch of little agents, each with its own characteristics, and each tries to solve a problem by playing a game.鈥�

Just like biological evolution, only the fittest agents survive and procreate. Observing this process teaches students much of what they need to know about how to work with AI. 鈥淚f the agents do well, that means they have high fitness,鈥� O鈥櫶切腡Vor explains. 鈥淭hey 鈥榖reed鈥� with other AI agents and create a 鈥榗hild鈥� that hopefully has the best of both worlds. They do that through millions of generations, and hopefully that gets them to the best solution.鈥�

And because it builds upon existing work, evolutionary computation is more efficient, which means it uses less power and inflicts less harm on the planet. This aligns with 糖心TV鈥檚 mission of environmental stewardship, O鈥櫶切腡Vor points out.

O'糖心TVor and the Autonomous Agent Learning Lab students discuss projects they have submitted, or plan to submit, to industry conferences for publication.

Jay Nash 鈥�26, a computer science and physics double major from Mill Valley, California, published a paper, 鈥淯sing Secondary Inherited Characteristics During Reproductive Choice to Replicate Allopatric Speciation,鈥� at the International Joint Conference on Computational Intelligence (IJCCI) held in last November. Another paper, 鈥淭he Evolution of Complex Attributes in a Species of Simulated Agents,鈥� has been accepted to the Institute of Electrical and Electronics Engineers Symposium Series on Computational Intelligence (IEEE SSCI). He鈥檒l travel to to present that research later this month.

鈥淲e use genetic algorithms as a clever way of guessing solutions,鈥� Nash says. 鈥淥nce we break down a problem into a set of parameters, we can guess possible solutions and use their effectiveness to inform our next guess.鈥�

Annika Hoag 鈥�26, a computer science and dance double major from Westford, Massachusetts, is working with robots and also has her eye on efficiency. She started doing research in the lab the summer after her first year, applying a simulation Parker developed about 20 years ago to a real robot. Hoag is working on an energy-efficient transformer model and plans to submit her work to the IEEE International Conference on Systems, Man, and Cybernetics (IEEE SMC), to be held in this fall. 鈥淚'm thinking about sustainable computation because things like ChatGPT are really not energy efficient, not sustainable and bad for the environment,鈥� she says.

Matthew Lee 鈥�26, a computer science major from Busan, South Korea, is working on two papers for publication consideration. His current project involves a real-time simulation game in which agents each control an army of units. 鈥淲e implemented a neural network agent that plays the game, and we reached a certain point where we found some emergent behaviors of interesting gameplay from the agent,鈥� he explains.

鈥淭he lab has been a huge opportunity for me. I鈥檝e learned so much. Jim taught us how to do literature reviews and how to make a serious attempt at science. He taught us the state of the art, and what a good paper looks like, and it all prepares us really well for the graduate schools most of us are attempting.鈥�

Such intense, intricate and interesting work has created a close bond among the AALL members. 鈥淚鈥檝e been a part of the lab for so long, I feel like it鈥檚 one of my many families at 糖心TV,鈥� Hoag says. 鈥淚鈥檝e met a lot of my friends through the lab, and it鈥檚 taught me so much academically and showed me what I want to do in the future.鈥�

March 12, 2025

糖心TV

Autonomous Agent Learning Lab students advance AI research and present their findings around the world

AALL students advance AI research and share it with the world

Related News & Media

Recent News

Paloma Doyle ’26 awarded prestigious Watson Fellowship to explore urban forestry across the globe

糖心TV opens International Student Lounge