Research and Discovery
A Summer of Data Science Homework?
Most people are familiar with artificial intelligence machines that learn from humans. Think Sophia, the humanoid robot, or Watson, the supercomputer that once competed on Jeopardy!. But what if AIs could learn from each other? What if they could learn to talk to each other, without human input? That’s what Visiting Assistant Professor of Computer Science David Perkins and his research group are exploring this summer.
The group that includes Sosina Abuhay ’23, Sam Aravilli ’23, Siqi Fang ’23, and Awildo Gutierrez ’23 has been meeting remotely to train liquid state machines (LSMs), a type of artificial intelligence, to precisely recognize various inputs. They now hope to begin training multiple LSMs to interact and communicate with one another to complete more complex tasks that a single LSM would not be able to complete on its own.
The simplest explanation of this type of code is that it recognizes and responds to patterns like a human brain and can be trained like a human brain. LSMs are special type of neural network, a set of algorithms based on brain structure that recognizes and responds to patterns. “You have to treat the liquid state machine like an organism ... It’s like you have this child that’s starting to learn something from scratch,” said Aravilli.
According to Perkins, most LSM research has focused on training the machines to function similar to other types of AI. His group’s research, which emphasizes communication between LSMs, thus addresses an under-examined potential. “We’re jumping in, trying to see if we can find a niche and contribute to this new field, which is only five to 10 years old,” Perkins said.
The students said that Perkins has encouraged them to guide and explore their research as they see fit. While the professor offers advice on the research process, the students have largely reviewed other studies, taught themselves a new coding language, written code, and run code on their own. “It feels like the professor is learning with us. It’s just that he’s a more experienced researcher,” Fang said.
“We were supposed to start early June, but they started meetings weeks early and were hooked,” Perkins said, describing the students’ enthusiasm for and commitment to the project. “All I did was give them some ideas that they might start with, not expecting them to start until later on … they’ve done 98 percent of the coding, planning, scheduling. I am amazed at how well this team is working together.”
Digital technologies and modes of thinking continue to change the world. Hamilton is responding by instilling among its students the skills to communicate and work effectively in this environment.
Teamwork has likewise been an essential part of the students’ learning process. “Usually, in the CS department, whenever we’re given projects, we do them by ourselves,” Abuhay noted. “Here the four of us working on a single code together, and every one of us comes with different ideas. Then we discuss how to approach a single problem and pick one that we think will get us to what we want to get.”
The students have enjoyed familiarizing themselves with both emerging technology and the computer science research process. “This summer confirmed my suspicions that research is one of the hardest things people do, but also a real, attention-gripping and meaningful experience,” Gutierrez said. “It’s worth the enormous amount of effort to feel that we are not only understanding the most current science, but also starting to improve on it and do things never done before.”
A Summer of Data Science Homework?
Wear Your Tech on Your Sleeve
In the spring semester computer science course “Wearable Tech,” students learned how to create fun, useful gadgets that could attach to a variety of items, from hoodies to baseball gloves to plants, to relay information and provide additional services.