Schmidt Futures conference at U of T builds global community of researchers advancing science with AI

Sep 27, 2023

Dr. Jessica Leivesley describes how she uses machine learning to make collecting ecological data more efficient.

Advances in machine learning are poised to revolutionize how we conduct scientific research, empowering us to address a range of grand challenges, from accelerating medical breakthroughs to developing new ways to address climate change. Yet, the coming shift requires a new generation of leaders—researchers with the vision and skills to develop an AI toolbox that accelerates science.

In mid-August, close to 120 of the world’s brightest minds working to advance this goal congregated in Toronto for a conference sponsored and co-hosted by Schmidt Futures. Participants exchanged ideas, forged connections and friendships, and shared their progress on projects harnessing machine learning across a dazzling array of fields and challenges.

This gathering of minds was the inaugural conference of the Eric and Wendy Schmidt AI in Science Postdoctoral Fellowships program. The prestigious program gives more than 100 fellows (including nine from U of T) access to training and mentoring, enabling them to incorporate cutting-edge machine learning techniques into their scientific practice. The goal is to create a global network of researchers capable of propelling science into the age of AI. At U of T, this program is co-led by Professor Lisa Strug, Director of the Data Sciences Institute and Professor Alán Aspuru-Guzik, Director of the Acceleration Consortium. The Vector Institute also supports the program.

“The Schmidt AI in Science Conference has been a remarkable opportunity for postdocs, faculty leads and program managers to gather and discuss their research and the future growth of this program,” said Professor Alán Aspuru-Guzik. “To have the first iteration of this event in Toronto and at U of T demonstrates that the Greater Toronto Area is the place to be in the field of AI for science.”

Schmidt Futures invited U of T to join the program in fall 2022, along with eight other universities in Singapore, the United Kingdom, and the United States. As the only participating Canadian university, U of T was honoured to co-host the program’s inaugural conference and to help establish a global community that supports and emboldens its first cohort of fellows.

Day three of the conference featured a spotlight on U of T’s Schmidt Fellows and the many ways U of T is working to adapt AI to drive scientific and medical breakthroughs. U of T staff led rotating groups of conference participants around the St. George campus to hear brief presentations from four U of T Schmidt Fellows on the work the program has enabled them to pursue.

“Our postdocs have immersed themselves in diverse, important projects, integrating novel AI tools to elevate their research goals,” says Professor Lisa Strug. “Having the opportunity to showcase their work amongst their peers and AI leaders in the field is an invaluable experience that will potentiate their research ever further. We are proud of what they have achieved since beginning these fellowships.”

The first stop on the tour was Acceleration Consortium director Aspuru-Guzik’s “self-driving” organic chemistry lab in the Lash Miller building, where postdoctoral researcher and Schmidt Fellow Felix Strieth-Kalthoff conducts his work. The room is full of industrial and scientific equipment stations (e.g., for mass spectrometry), each connected by a complex array of spaghetti-thin tubes carrying liquid solutions between them. The centrepiece is a platform where a small robot automates the manual labour of mixing, pouring and sampling that organic chemistry research requires.

Strieth-Kalthoff led conference participants around the packed room as he explained and answered questions about how this AI-guided system, which can run autonomously for days, has eliminated massive amounts of manual labour and sped up his research. His Schmidt Futures project explores novel ways of using chemical synthesis and catalysis to discover new drugs and other functional materials. Strieth-Kalthoff envisions increasing the system’s efficiency by automating the task of purifying solutions, for example.

To have the first iteration of The Schmidt AI in Science Conference in Toronto and at U of T demonstrates that the Greater Toronto Area is the place to be in the field of AI for science.

In the Myhal Centre for Engineering Innovation & Entrepreneurship, the tour was treated to an overview of Fatema Tuz Zohora’s research in computational biology. As a U of T Schmidt Fellow and postdoctoral researcher at the Princess Margaret Cancer Center’s Schwartz lab, Zohora is investigating new ways to reduce anti-cancer drug resistance implicated in nearly 90 per cent of cancer-related deaths.

To have the first iteration of The Schmidt AI in Science Conference in Toronto and at U of T demonstrates that the Greater Toronto Area is the place to be in the field of AI for science.

One likely cause for this resistance lies in how cells communicate with each other. To explore this possibility, Zohora combines computer vision, natural language processing and other machine learning techniques to analyze data from multiple molecular layers of cell-to-cell communication. Her groundbreaking work could lead to new forms of therapy that make life-saving cancer drugs more effective.

Just down the hall from Zohora’s talk, Jessica Leivesley, a postdoctoral researcher in the Department of Statistical Sciences, described a technique she developed to make gathering ecological data more rapid and efficient. Her research as a Schmidt Fellow combines machine learning techniques to automate how scientists monitor fish species populations—a vital task in managing freshwater ecosystems.

Typical sampling methods are invasive or even lethal for fish. Leivesley’s non-invasive method identifies smallmouth bass and other fish species by analyzing how sound waves move through lakes. As the sound waves pass through the bodies of fish, they reflect back differently depending on the size and shape of the fish’s air bladder. By tracking these subtle differences in the hydroacoustic data, Leivesley’s machine learning techniques can identify the presence of different fish species. Her approach could transform how ecologists gather critical data and supercharge their ability to track environmental change.

Our postdocs have immersed themselves in diverse, important projects, integrating novel AI tools to elevate their research goals.

The final stop on the tour took participants to the new home of U of T’s Data Sciences Institute at College St. and University Avenue. Here, Soukayna Mouatadid, a U of T Schmidt Fellow and postdoctoral researcher in U of T’s Department of Computer Science, described her work incorporating machine learning tools in climate modelling. Her work demonstrates how we can use these tools to develop a new generation of meteorological models that make more accurate predictions, even as climate change increases devastating weather events such as heatwaves, drought and floods.

Our postdocs have immersed themselves in diverse, important projects, integrating novel AI tools to elevate their research goals.

Mouatadid focuses on sub-seasonal forecasting, which predicts weather patterns between two and six weeks in advance, using a different mixture of data and variables than regular, near-term forecasts. The accuracy of current sub-seasonal forecasting predictions is relatively poor, yet this time window is critical for preparing for extreme weather events. Mouatadid recognized that current approaches could be improved by using historical data better. To address this challenge, she uses machine learning to track invisible patterns in 40 years of data across 14 meteorological variables. She recently won an award for her work from the US Bureau of Reclamation and the National Oceanic and Atmospheric Administration, and she is already developing a more granular model capable of even more accurate predictions.

The Eric and Wendy Schmidt AI in Science Postdoctoral Fellowship program has not only supported the work of these stellar researchers at U of T, but also folded them into a global network of like-minded scholars doing cutting-edge work. As this Schmidt Futures program attracts new cohorts of talented researchers, it will continue to propel the AI revolution in science.

Applications for the second cohort of Schmidt AI in Science Postdoctoral Fellows at U of T are currently being accepted until October 2, 2023.

Apply here