How do you become a U of T student researcher in machine learning?

(Originally published on The Varsity)

Seven U of T students discussed their experiences and insights as machine learning researchers at the AI Student Panel Talk. The talk was organized by the AI Squared Forum, and took place in the Bahen Centre of Information Technology in September.

At the panel, the speakers answered moderator-curated questions and questions from the audience, with topics ranging from learning resources, personal development, and time management, to advice for making the first step in research.

What skills and personal qualities are necessary?

The speakers shared the skills and qualities they believe are important to become a good student researcher. Their answers can be grouped into three main categories: personality, technical skills, and social skills.

Kyle Hsu, an undergraduate student studying engineering science who is researching at the Vector Institute for Artificial Intelligence, believed resilience and self-confidence are key qualities to have. Knowing programming tools and languages, as well as abstract symbol manipulation, is also crucial to understand machine learning research.

Since most research projects cannot be done without cooperation, student researchers also need to develop social and communication skills.

Where can students get hands-on experiences?

Winnie Xu, a third-year undergraduate student researching robotics at the Vector Institute, said that she learns with online tutorials. The speakers also recommended participating in Kaggle competitions and Hackathons, as completing a project from start to end is a valuable experience, both in learning technical skills and in project management.

While it can be difficult to finish a project, Jacob Kelly, a third-year undergraduate student conducting research at the Vector Institute, recommended joining topical student clubs as a way to learn together and find a supportive community. In addition, Hsu mentioned that reproducing a published research paper is another great way to develop a deeper understanding of the topic students may be interested in.

According to the panelists, the best ways to stay up-to-date with the latest news of machine learning research include following active researchers on Twitter, subscribing to technical blogs, and auditing graduate-level machine learning courses.

Massive Open Online Courses (MOOCs) are also helpful resources. A question from an audience member concerned how the speakers got into machine learning in the first place. MOOCs were mentioned again and again, so the moderator asked the panel: “Who took a MOOC when you got started?”

Five out of seven speakers raised their hands.

How can students get their first research opportunity?

For students looking to get involved in research, finding the first opportunity is often the most difficult. Getting a personal connection could make the process much easier, such as by reaching out to alumni.

But most speakers on the panel started with cold-emailing professors; some sent up to 100 emails trying to find a position, and a few got a reply after one or two months. The speakers also suggested strategies for cold emailing.

For example, applicants should tailor their emails to specific professors and mention why they think their research is interesting. The panelists also recommended students to attach a resume, list all the relevant courses and projects, and keep the email short - usually no more than a couple paragraphs.