It’s a serious study, with each AALL student committing to AALL for about 30 hours a week, in addition to class work and other extracurricular activities. However, students say the knowledge they gained and the opportunity to publish at professional conferences as undergraduates is worth the investment.
“The lab is very supportive. says Delin Götzgin ’27, a dual major in computer science and statistics and data science at Izmir, Turkey. “We are working at a graduation level while we are undergraduate students. After we finish our project, our ultimate goal is to contribute to science. By getting the publication, we can do this. If you want to apply for graduate school, it is important that we be published as we show that we can commit to and finish a new research project.”
AALL and other Computer Science Labs are funded by Honorary Trustee Jean C. Tempel ’65, and have been teaching at Conn since 1999 and overseen by Professor Gary Parker of Computer Science whose research focuses on the methodology of learning for autonomous agents. Since it was founded shortly after Parker arrived, AALL has continued to grow, with about 20 students currently working on projects of about 20 at any time of the school year.
“It is important to create an environment where students can learn problem-solving techniques that will help them move the thoughtful use of established applications beyond the realm of scientific discovery,” says Parker. The professor is the student’s primary advisor, and O’Connor guides them through Parker’s concepts and ideas.
As AALL’s research focuses on evolutionary calculations, instead of creating AI models from scratch, students evolve existing models. “Playing games like chess and some video games is a good example of how to push the limits of AI. “We create small agents each with their own unique characteristics, each one trying to play the game and solve the problem.”
Just like in biological evolution, only suitable agents survive and reproduce. Observing this process teaches students many of the things they need to know about how to collaborate with AI. “If the agents work out well, that means they have high fitness,” explains O’Connor. They “bred” with other AI agents and create “children” who have the best world of both. They do it through millions of years of generations, hopefully it will lead them to the best solution. ”
Also, since it is based on existing work, evolutionary calculations are more efficient. This means that it uses electricity and is less harmful to the planet. This is consistent with Conn’s environmental management mission, O’Connor points out.
