The end of every academic year brings around many recurring activities. Exams are one such example, and final project presentations are another classic year-end hoop for students to jump through. But on March 28, in the upper mezzanine of the Trottier Building, a group of 23 students held their project presentations, with an interesting twist: they had been entirely instructed by other students over the course on an eight-week boot camp.
Not only were participants instructed by their peers, the topic area was nothing less than machine learning, which is an application of artificial intelligence (AI).
Jad Hamdan was one presenter. The Computer Science and Mathematics student’s work was a two-part project to turn musical notation into a code that a machine could understand; and train the machine to generate music using this transcribed notation. The net result came out like a bit of ragtime piano, a classical music piece with blue notes.
Commenting on the experience, Hamdan said, “The sheer amount of material covered in the boot camp combined with the mentorship and advice from upper-year students made for one of the most enriching experiences I’ve had at McGill.”
Fellow boot camper and Mathematics/Computer Science major Rosie Zhao demonstrated her work on artistic neural style transfer, a process that takes as input two images and uses optimization to generate an image that emulates the style of one image while retaining the content from another. The technique has strong application in the visual arts industry. She was also enthusiastic about her journey: “I have a deeper passion for machine learning and I believe that the boot-camp has equipped me with the skills to navigate this field further,” she said.
Yet another presenter Gregory Pope created a system that had been taught to identify the qualities that made up a good wine. Given the right data inputs – which were admittedly more complex than just tasting it – the machine could parse the chemical analysis and determine which wines were great versus merely average or inferior.
The project only examined the red and white versions of Portuguese Vinho Verde wine, however, so it will be some time before people can use their phone on their next trip to the SAQ. But that was not the point, nor was it the goal of any of the ventures on display. “This project was a fantastic opportunity to explore the world of machine learning, and I was introduced to many topics I might not otherwise study until third or fourth year,” said Pope, who is also a Computer Science student.
No experience necessary
Although there was an extraordinary diversity in terms of the potential applications of the various projects, what united each of them – and what made the whole exercise so remarkable – was that prior to taking the course, not a single student had any previous experience with machine learning. And yet, in the short space of a few months, and on top of their normal course load, these students were each able to fashion their own intelligent machines.
The boot camp – the official name is MAIS 202 – was the brainchild of the McGill Artificial Intelligence Society (MAIS), a student club run out of the Faculty of Engineering. The 13 members of the club’s executive committee wanted first- and second-year students who didn’t have the necessary prerequisites to take the upper-year classes on machine learning to have the opportunity to learn about the subject.
“As upper-year students, we’ve taken all the courses already, and we know what gaps there are,” explained John Wu (Beng ’19), who co-founded MAIS in 2017.
Lectures, assignments… and snacks
Taking cues from University of California Berkeley’s DeCal program (where students create and facilitate their own classes on a variety of subjects) Wu and other executive members of MAIS set up an eight-week program with lectures given by fourth-year students, weekly assignments and a final project. Existing MAIS sponsors kicked in some snacks to keep the students’ energy levels up, and some professors supplied classroom space.
The boot camp was open to students from any Faculty were accepted into the program, so long as they had some mathematics and coding background (which was tested via an entrance exam). Perhaps the most important ingredient was that they had to show they were passionate about the topic.
“As there were mostly engineering students who applied, some of them may not be so good at expressing themselves,” Wu said. “But when they have a passion for something, you can tell.”
The selection process was on target, as there was plenty of passion on display in the Trottier lobby during the project presentations, including among the student-instructors.
Wu graduates at the end of May and will go on to a job at Microsoft, but he is confident his successors at MAIS will carry on.
“Based on the kinds of requests we’ve received, we are thinking of adding something on ethics in AI,” he said speaking of the future of the outfit he helped to launch. “There’s so much interest out there, and AI is something that people just don’t understand well enough – we need to be able to explain it better to the world.”