Purveyors of fake news had better beware: A machine-learning engineer who took up the challenge of sorting the false from the true has created an artificial intelligence-driven program to defeat you, and won a $1-million prize in the bargain.
Bill (Yu) Wu was announced Thursday as having bested a field of 150 programmers, designers and engineers who took up the challenge posed by the Leaders Prize, first revealed at Communitech’s signature tech-for-good event, the True North Festival, in 2018. At the 2019 Festival, the specific challenge was issued: Find a way to use AI to solve the global problem of fake news and misinformation that has been undermining democracies.
To inaugurate the challenge, Communitech partnered with the Schulich Foundation and the Leaders Fund, along with the University of Waterloo as an academic partner. A team of judges drawn from the fields of data science, analytics and business reduced the initial entries down to nine finalists.
The winner was intended to be revealed at True North 2020, which was cancelled due to the COVID-19 pandemic, a global health crisis that itself has generated a fair amount of misinformation.
So instead, the winner was revealed on True North TV, the YouTube series that has been delivering True North’s message during the pandemic, with the announcement Thursday by Communitech CEO Iain Klugman and Leaders Fund co-founder David Stein.
Klugman said, “While we anticipated that the Leaders Prize would be a great opportunity to showcase Canada’s leadership in AI, we did not expect that the world would face a global pandemic – and a flood of related fake news – in the final stretch of the competition.
“The calibre of submissions and remarkable talent atop the Leaders Prize scoreboard truly exceeded our expectations and made us even more proud of the Canadian tech community.”
In the first phase of the challenge, teams from across Canada submitted solutions that automatically verified a series of claims, flagging whether they were true, partly true or false, in addition to providing supporting evidence for that conclusion. In Phase 2, the solutions also had to provide evidence to support the truth rating and text explanation for each claim. The top prize was then awarded to the most effective AI-based fact-checking solution.
For Bill Wu himself, Thursday was surreal. “It feels kind of unreal to me now. It hasn’t sunken in yet.”
Wu, a University of Toronto grad with a Bachelor of Applied Science in Electrical Engineering, is the sole member of Team Ava, whose product – Ava – is named after the principal character in the 2014 sci-fi thriller Ex Machina.
After graduating in 2017, Wu began work as a machine learning engineer at a Toronto startup. A friend told him about the challenge posed at True North in the summer of 2019, and Wu began exploring the possibilities. He worked nights and weekends and, earlier this year, took some leave from his Toronto job to work on the problem full-time.
Wu said Ava automates the work of fact-checkers, by taking in as an input “a claim made by someone, and as an output, it will tell you either if the claim is true or false or partly true. It will also give you articles to support that rating, and then offers a short explanation on why the claim is rated as it is.”
Wu says social media companies and non-profit agencies hire humans to do this, but says that “I don’t think Ava will replace humans . . . with the current status of AI technology, there are many things that humans can do better than computers. So it’s not about replacement: it’s more about augmentation and freeing up of their time to do the more complicated tasks.”
Wu urges designers to not give up if they encounter obstacles. Some 50 times, he encountered something he thought would work, but didn’t.
“Do not get discouraged when you have a lot of things that don’t work out, especially when you have multiple ones in a row. Especially at the start and middle stages of the project where I had all these ideas where I thought, ‘That’s really going to work, that’s definitely going to help,’ and it just didn’t. And there are some things where it was a long shot, but I thought, ‘I’ll try it,’ and it worked. So it’s important to keep an open mind and just keep trying.”
Ava is not yet ready for the market. There’s a lot of infrastructure that needs to be in place to support the core program. Right now, Wu is not sure what’s next for him.
“I don’t have the money right now. And where do I go from here? I think I’ve always wanted to work on something that is at the intersection of what I’m good at and what I find interesting and what adds value in this world.
“This doesn’t really change that. More specifically in terms of AI – how does intelligence work, how do humans do the things that are so hard to replicate? My interest doesn’t change. I will continue to try to work on these projects.
“The exact form of it, I’m not sure. The money does give me some more flexibility in how I do things.”
Wu sent a shoutout to those who helped him along the way: “I have many colleagues who have been very supportive and capable and intelligent, and it’s through working with these people that I also gained in skill, and over time, became better and better at doing machine learning-based software.”
And he appreciates the work done by the organizers of the Leaders Prize: “Thanks to the organizers and the partners that provided a platform for this challenge and for the money.”
Wu says the Leaders Prize is “. . . a way for people to come forward and try out their skills at a challenge that is meaningful. I think sometimes that people don’t know what to work on.”
At the Thursday announcement, David Stein said, "We look forward to getting this solution into the marketplace where it can play an important role in combating the spread of fake news.”
After the announcement, Feridun Hamdullahpur, President and Vice-Chancellor of the University of Waterloo, said the victory for Wu marks an exciting milestone not only for the first Leaders Prize, but for the nation. “Throughout this competition, it has become abundantly clear that Canada’s emerging leaders in technology and innovation are also poised to lead the world in many other important ways – shaping the future of information sharing, data, and democracy itself.”
In addition to the founding partners, the Leaders Prize competition was supported by Communitech partners CIBC, Cisco, Desjardins, Politifact, Thomson Reuters, Wawanesa, Borealis AI and the Vector Institute.
The other teams on the short list included: Jaymody, Supersonic, dbrait2000, NL Philosophers, Traveling Wilburys, ImAnLearn, Nima-Fit, and ConvolutedLP
More information on the competition can be found at: Leadersprize.ca / Prixleaders.ca.