What we learned at ElevateAI
Wednesday was an exciting day for Canadian AI. Leaders in the Toronto tech community organized our energetic ecosystem to come together and discuss the real, buzzing potential — and perhaps ethical responsibility — of doing what it takes to claim our place as a global leader in AI. Many heartfelt thanks and congratulations to Razor Suleman, Candice Faktor, Darin Graham, the Vector Institute, and MaRS for making the event a resounding success.
We were proud to sponsor the AI track and give our team the opportunity to share our values and some insights on our approach to applying AI to help companies maximize mutual lifetime value with customers. But we were most excited to hear from others, to build a few more bridges that, compounded across our community, might lead to 10 more Shopify platform companies we know we can create. To continue the momentum, we thought we’d share some of our key takeaways from the event.
Dan Debow, CEO of team communication platform Helpful.com, closed the day with an inspiring speech listing factors we need to consider to realize Canada’s AI potential. As each item in his list telescoped points others made throughout the day, we thought we’d sample a few to recap the event and guide our collective efforts going forward.
Curiosity-Based Research
Many presenters, from long-time deep learning researchers like Geoff Hinton and Yoshua Bengio to younger pillars of the community like Tomi Poutanen of Layer6 AI, underscored the importance of curiosity-based research to fuel the future economy. It’s largely due to intellectual freedom that Canadian research labs are leading the way exploring formal, mathematical representation of intelligence. And we have to get comfortable with the fact that researchers may not predict the most impactful commercial value of their diligent quest. Geoffrey Hinton likely did not know that his seminal work in backpropogation — which helps train the many parameters of a deep neural network — would eventually led to capabilities like language translation at scale. The team at D-Wave still don’t know what products their delicately entangled qubits may enable: we look forward to exploring how the Quantum Machine Learning cohort at the Creative Destruction Lab will apply this entirely new approach to optimization in machine learning this fall. The tip of the innovation spear, this curiosity must remain unconstrained even if it creates many efforts that never go commercial. It’s up to the entrepreneurs — the visionaries — to unlock the value of a capability down the line.
Data
Canada can’t get around the fact that it has less people than China and the US, but it can level the playing field by stepping up efforts to make data open and accessible to researchers. Efforts like those at ThinkData Works to curate and supply open data sets can only support future Canadian research. Professor Yoshua Bengio pointed out that much of his research can be performed on “toy” data sets, given that long-term research questions focus on exploring the possibilities of algorithms, not using an algorithm to solve a problem within the constraints and realities of a real set of enterprise data. The difference in mindset and approach should not go overlooked as enterprises seek to apply AI and vet talent that can have business impact.
Experiments in Social Policy
Panelists inevitably discussed fear of automation and workplace displacement due to AI (although we think this topic often distracts us from more pressing near-term adoption issues). Gary Bolles mentioned our near-sighted bias and lack of creativity in rethinking what work might look like as parts of tasks can be automated. Like Andrew Ng, he believes the workforce of the future must constantly retrain and adapt their skills to keep up with the rapidly changing technology. We agree, and are therefore proud to support the wonderful Ladies Learning Code team at their fifth annual National Learn to Code Day September 23. Alongside automation, participants also mentioned concerns about explainability and fair use of AI systems. We agree with Yoshua Bengio’s response that it’s important the community appreciate that the models powering AI are nonlinear and should be respected as such. We shared our thoughts on regulatory, conceptual, and technical approaches to explainability and bias at a recent NextAI seminar at Rotman, as well as our credo on building products that respect users’ goals at Wrangle in San Francisco.
Building a Community
Finally, Debow and our CEO & Founder Steve Irvine both underscored the importance of openness, boldness, and mutual support in continuing to push the momentum in our thriving AI ecosystem. To win, we don’t have time to waste on conflict or paranoid competition. We want healthy competition so we build quality products, but we want to learn from each other. We have to lean on veterans like Steve Woods at Nudge.ai for their wisdom on technical architecture. We have to catch the infectious energy Jodi Kovitz breathes to support women in tech. We have to be open to entirely new ways of working with academia like the team at Element AI. And we have to see ourselves as a community, driven by our feeling that we’re doing something special at a turning point in history.
