Supercomputers played a key role in the scientific discoveries recognized with recent Nobel prizes in physics and chemistry, such as Geoffrey Hinton's work on artificial neural networks and the insights into protein structures from the team at Google's DeepMind.
Rapid processing of massive datasets and complex calculations are among the capabilities that make supercomputers valuable tools for tackling hard-to-solve challenges. By enabling significant breakthroughs in areas like resource and energy management, weather and climate analysis, disease and treatment modelling, and material design, for example, supercomputing has already left an outsized mark on society. Now, increasing adoption of artificial intelligence (AI) applications has led to a further spike in interest in supercomputers.
"As more people interact with AI, there is more awareness about the impact of supercomputers on everyone's lives," says Ryan Grant (Electrical and Computer Engineering), whose goal at the Computing at Extreme Scale Advanced Research (CAESAR) lab is to unlock the full potential of supercomputers in a Canadian context. "Many people have used ChatGPT, which was trained on a supercomputer. The two-metre COVID-19 distancing rule also came from a supercomputer."
The surge of interest in AI - and its anticipated economic benefits - has helped swell the numbers of people working in supercomputing across the globe. As countries seek to position themselves as technology leaders, many take a three-pronged approach consisting of a "commercial cloud, scientific research supercomputers and industry-driven innovation supercomputers," says Dr. Grant, who compares this to "three legs of a stool."
Canada has the first two components. "One, we have cloud computing, which is used for many separate, fully independent tasks, such as streaming Netflix or using ChatGPT, although large jobs like training AI are very expensive," he explains. "Two, we have research supercomputer infrastructure under the stewardship of the Digital Research Alliance of Canada, which is absolutely critical to scientific discovery and learning."
An additional component - industry-driven research supercomputers - can provide much-needed stability, as does a third leg of a stool, proposes Dr. Grant.
"When we place supercomputers at the intersection of academia and industry, this creates a research pipeline where scientists and industry experts work collaboratively on solutions."
- Dr. Ryan Grant, Smith Engineering at Queen's University
Due to their proximity for taking ideas to impact, these public-private partnership models are known as "innovation supercomputing," and Dr. Grant suggests looking at international examples - such as the national laboratories in the U.S. and the Barcelona Supercomputing Center, the CSCS Swiss National Supercomputing Centre, and the Leonardo supercomputer in Europe - for insights on how to catalyze success.
Studies show an impressive economic impact, with a $43 return for every dollar invested in the U.S. In the E.U., where job creation is considered a factor, the ratio is estimated at 60:1. What's more, in a recent report from Europe, users of such systems credit these collaborations with leading to the "top three significant discoveries for their business area in the last 10 years," says Dr. Grant. "When industry partners put efforts into supercomputing projects, they tend to see revolutionary discoveries."
Supercomputers played a key role in the scientific discoveries recognized with recent Nobel prizes in physics and chemistry, such as Geoffrey Hinton's work on artificial neural networks and the insights into protein structures from the team at Google's DeepMind.
Beyond industry success, the innovation supercomputer model accommodates the "rapid deployment, maintenance, and updating of infrastructure," due to funding support from partners, Dr. Grant notes. "It's a win/win for everybody. Industry gets inexpensive access to research and innovation, and the taxpayer, who foots the initial bill, sees economic returns from a system that is self-perpetuating."
Ryan Grant is leading efforts at Queen's to build supercomputer infrastructure, including an advanced test bed system for testing and deploying brand new technology. This will create a foundation from which Canada can assume a leadership role in supercomputing.
Insights from international examples can inform a Canadian model, says Dr. Grant, who gained experience at Sandia National Laboratories, a leading supercomputer facility in the U.S., and from collaborating with the international supercomputing community.
"If Canada wants great, secure, sovereign supercomputing, it needs to play the long game and develop talent and test beds for technology development," he explains. "Universities can play an important role in providing a foundation, from where can layer on research focused on giving Canadian firms, and especially startups, a competitive advantage."
Dr. Grant is already making headway on the talent aspect: with 20 supercomputing experts, his lab is now one of the largest in the world focused on supercomputing architecture, and specifically software for supercomputing networks.
"We're working on having a system up and running," he says. "We're also building support for a larger innovation supercomputer, and we've consulted with over 90 different institutions and companies to understand their needs."
Another focus is on "creating an advanced test bed system, where we can deploy and test brand new technology," explains Dr. Grant. "This will allow us to stay on the leading edge of tech and inform our long-term outlook."
Building infrastructure and training experts can allow "Canada to become a global leader in supercomputing - and see significant economic benefits," he says, proposing to leverage a number of Canadian advantages, such as the Pan-Canadian AI Strategy, world-renowned experts in AI, and a recent funding commitment of $2.4-billion for AI infrastructure announced in the 2024 federal budget.
While Canada is a leader in AI, its most powerful supercomputer only ranks 123rd in the world, and Dr. Grant emphasizes that "having an innovation supercomputer that can serve as an AI factory will supercharge AI as well as other areas of the economy."
Boosting Canada's capabilities to compute securely as well as "know how to build and use these systems will allow us to catch up to other nations around the world," he says.
"We need to develop these capabilities here at home, so we can create a uniquely Canadian infrastructure to ensure Canadian businesses and universities have access to world-class experts. This will accelerate both research and business throughout Canada."
To learn more about Dr. Grant's research, read "Sparking Canada's supercomputing renaissance" in the Queen's Gazette.
