Mind and machine
Ajay Agrawal, the man behind the programme that is home to the greatest concentration of AI start-ups in the world, tells Indira Kannan why he thinks machine learning will drive the future
The president of Alibaba came from China; the governor of the Bank of England came from London; professors arrived from top US universities like Harvard and the Massachusetts Institute of Technology; tech entrepreneurs and scientists travelled from Silicon Valley; and business leaders flew in from across the world. The event that drew them to the University of Toronto’s Creative Destruction Lab, or CDL, last month was the fourth annual “Machine Learning and Market for Intelligence Conference”.
As speakers and attendees interacted and networked at the day-long conference on the future of Artificial Intelligence and its role in the economy, Ajay Agrawal, the Indo- Canadian founder of CDL, marvelled yet again at the transformative power of AI. Alibaba’s was a case in point, which he learnt about while chatting with Michael Evans, president of the Chinese e-commerce giant. When Alibaba’s drivers go out to deliver packages, they must first call the customers to arrange the details of the drop. Since Alibaba started using AI to make those calls, Evans told Agrawal, the company had been saving a staggering 160,000 hours of labour every day.
In recent years, Toronto has emerged as an important hub for AI. And Agrawal is one of the drivers of that growth. Professor of entrepreneurship at the University of Toronto’s Rotman School of Management, Agrawal is also the co-founder of Kindred AI, a company that builds machines with human-like intelligence and was 29th on MIT Technology Review’s 2017 list of 50 smartest companies in the world.
Agrawal, 48, had always been interested in science fiction and AI, but when he founded CDL six years ago, it was as a “seedstage programme for massively scalable, science-based ventures”, as it describes itself. “The lab was open to inventions from around the world and in any discipline,”
Agrawal says. But then he started noticing a trend. A student at the University of Toronto approached the lab saying he had developed a technique to predict which molecules would most effectively bind with which proteins in a way that would help transform drug discovery. Another graduate student said he had used the same technique to look at video images from a car assembly plant to predict which cars had defects before they rolled off the assembly line. Next was a student who was using the technique to look at pixels in a medical image to predict which tumours were malignant and which benign.
“This started happening at a faster and faster pace,” Agrawal recalls. “And it quickly became apparent to us that this was unusual — we were still getting applications from biology and chemistry and so on, but so many were using the same technique of machine learning for such a wide variety of applications.”
This prompted Agrawal to set up the first specialised stream within CDL, exclusively for AI, supporting 25 startups working in this field. He had spotted the trend early — so early that he faced plenty of scepticism over the focus on AI. “People said there were not enough companies, there won’t be enough demand, there won’t be enough investor interest to have all 25 in one category. But of course it turned out to be just the tip of the iceberg. There was much, much more demand than that,” Agrawal says.
Today, CDL is home to the greatest concentration of AI start-ups in the world. What made this possible was also the local presence of world-class research support. “We build streams where we have people who are not just the best in Toronto or in Canada, but the best in the world in their field. And in AI we had that, so that’s why we started it,” Agrawal explains. “The computer science department at the University of Toronto and what is now known as the Vector Institute were critical as research hubs so that we have excellence in the root discipline.”
CDL has since expanded to Vancouver, Montreal, Calgary, Halifax and New York, and added more streams: Blockchain-AI, Cities, Energy, Health, Quantum, and most recently, Space. The companies incubated at CDL have generated over Canadian $3.1 billion in equity value. They are spread worldwide and include Atomwise, Heuritech, Automat and Deep Genomics. This year, for the first time, an Indian company, Bengaluru-based Nopo, entered the CDL programme in the Space stream. And in October, the Canadian government announced an investment of $25 million into CDL.
Agrawal continues to focus on the economics of AI in his research and practical work. This year, he co-authored with his fellow Rotman School professors, Joshua Gans and Avi Goldfarb, a best-selling book titled Prediction Machines: The Simple Economics of Artificial Intelligence (Harvard Business Review Press). The book seeks to explain the role of AI in businesses as a function that dramatically lowers the cost of prediction — a key input in decision-making along with judgement.
Agrawal was born in Vancouver to a Scottish- Canadian mother and an Indian father who had moved to Canada from Bhopal, Madhya Pradesh. He attended school in Vancouver and went to the University of British Columbia for his bachelor’s and master’s degrees as well as his PhD, apart from being a visiting scholar at MIT and the London Business School. He is also a co-founder of NEXT Canada, a non-profit group that works towards promoting technology adoption and helps entrepreneurs. He says he visits India occasionally to meet relatives but has no professional commitments in the country at present.
Ask him about its potential applications in India and Agrawal says AI offers the Indian economy either great potential or great risk. He cites the example of China, noting that the potential impact of industrial robotics was greatest there as the country was home to the largest amount of manufacturing that could use this technology. Similarly, the service sector, where India is a leader, is a field where AI could be put to maximum use. India could use its leverage as the home country of much of the data needed to train AIs in applications like chatbots that can be used in call centres. If it doesn’t, he cautions, India risks losing a lot of service sector business to countries that can develop AI to replace or complement many functions.
Apart from further research on the economics of AI, Agrawal is also working with companies developing AI embedded in two kinds of robots: “Industrial robots that are working in fulfilment centres for people who order goods and services from e-commerce companies; and human-like robots that can interact with humans in a way that makes people more comfortable talking to and empathising with machines”.
According to CDL, AI will drive over $15.7 trillion of growth in the global economy by 2030. Platforms like CDL are at the vanguard of that development. As Agrawal says about the coming impact of AI: “In some sense, that’s like asking which sectors in the early 1900s would be most impacted by electricity.”