I was in Toronto from March 22–23 to attend two conferences: Future Now, an open Microsoft event to highlight recent advances in AI; and AI + Public Policy: Understanding the Shift, an invitation-only event by Brookfield Institute for Innovation and Entrepreneurship and the Government of Ontario. Here are my notes from the trip.
Meeting with Elissa Strome, Executive Director of the Pan-Canadian AI Strategy, CIFAR
This was more just a meet-and-greet and an opportunity for me to go over some of the work we’re doing and to make sure that Canadian Institute for Advanced Research (CIFAR) is engaged with it as much as possible. CIFAR was pivotal in helping steer Canada’s AI research decades ago, and it’s going to be pivotal in helping us understand the effects of AI and society. It will be important to connect key academics and industry with policymakers in an ongoing fashion so that as the technology evolves, we can rapidly adapt to that change. As the guy in charge of understanding how AI is going to change the way the government provides policy options and services, I’m really looking forward to working closely with Dr. Strome and her team, especially on international standards and our key policy artifacts over the next couple of years.
One big takeaway: Dr. Strome suggested that the Algorithmic Ethics Advisory Board that I proposed be expanded to include a potential review of the science used by some of our projects as well. This way, departments would get advice from experts on the methodology used by their automation as well, which of course could have ethical considerations too.
Future Now, Microsoft
The primary purpose of attending this conference was to sit on a panel with some luminaries in the Canadian AI space to talk about the social impact of the technology. But I’ll get to that in a moment. I have to be honest, I skipped a lot of the conference by accident even though I had planned on attending most of it. When I went to put down my stuff in the green room, I had a chance to meet Mark Skilton. We got to talking for about an hour and so I completely missed the afternoon presentations. It was a fascinating discussion about the nature of working, whether a five-day work week is even necessary in the automated economy, the changing nature of higher education, etc. I think I’ll buy a couple of his books for the office because they’re quite relevant to a variety of work that we do.
I had the pleasure of appearing on a panel moderated by Andrée Gagnon from Microsoft, and including Mohamed Musbah from Microsoft Research Montréal, Kathryn Hume from Integrate.ai, and Khalid Al-Kofani from Thomson Reuters. I was happy to be able to bring a little bit of public policy perspective to an otherwise corporate-focused conference.
One big takeaway: for Canada, and the world, to navigate the coming shift, we have to be open to setting aside sacred cows for new solutions. Kathryn definitely had the most powerful point of the panel, which was that we have to move from focusing on hyper specialized training early on in education and career and instead focus on training people to just be better people. It will be soft skills like adaptability, empathy, and critical thinking that will make someone competitive in the future. There’s a large focus on training people solely to move into STEM fields, when probably the most adaptable people will be those with multidisciplinary backgrounds (like her).
The AI Shift, Brookfield IIE and Government of Ontario
It’s really hard to summarize the whole day into a couple of paragraphs. The AI Shift was one of the richest conferences on the topic of AI and public policy that I have been to. Personally, I feel like it would have benefited from a second day where we had the opportunity to dive deeply into some of the issues that were raised, because there was a lot of ground to cover in a short period of time. Brookfield are going to make most of the talks available, and include a summary report which will be very helpful helpful.
Literally every talk was a highlight for me and will be worth watching. The fireside chat between Taylor Owen and Ben Scott on the threat to democracy posed by AI media content recommendations was particularly amazing/wildly depressing. That said, probably the one that will stay with me was a 10-minute primer on privacy issues of AI in government by Lex Gill, a research fellow at the Citizen Lab. It was an intelligent, articulate, and passionate call to action for governments to understand the gravity of using these tools for socioeconomic or justice applications. I’m going to send it to every senior executive that I’ve heard speak about the need to test AI to inform policy.
Only myself and Brian Purcell at IBM spoke directly to public sector use cases. I really wanted to reinforce that AI for the public sector doesn’t necessarily mean high risk stuff like predictive policing. Even “dull” cases like the use of AI for information management or summarization of large bodies of text would be enormously useful to making government a more effective and efficient machine.
- The technical experts who spoke definitely reinforced that this is a shift that is happening very quickly. Even comparisons between capabilities in 2012 vs today we’re pretty eye-opening.
- There was a bit of pessimism that governments will be able to manage the shift, both internally, and with the public policy repercussions of AI. As the only Fed in the room, I took this advice to heart.
- If you take item one from above and multiply it by item number two, what you get is a little bit of trouble. It’s clear that if we try to manage this as a typical file, we will fail.
- Governments tend to have more data points on people who are more vulnerable, because people who are more vulnerable use services more often. There was some concern that government use of AI can create self-fulfilling feedback loops.
- Many people saw the need for algorithmic interpretability and accountability as an opportunity for government to improve its decision-making practices rather than a challenge. I was really happy to see how optimistic that perspective was!
- Generative modelling is an amazing emerging frontier — Graham Taylor showed some interesting examples of AI creating images of “new” faces based off of a training set of celebrities. There is so much to unpack here, but what drew me in were the intellectual property considerations. If I were to write a model that created new imagery, ostensibly I would own the output, but would I also be entitled to own all potential outputs given the training set and model as well? That could be incredibly stifling on innovation.
- Don’t treat consultations like boxes to be checked. Instead, love them. Love the challenge, the lessons, the perspective. Love the chaos of democracy. Every push, contribution, and negotiation has made me a better public servant and a better human being. Interacting with stakeholder groups has become my favourite part of the job.
- The public service has long lost its monopoly on policy design and advice. Our job is now to convene a discussion of opinions and draw a line of best fit. Think-tanks, academia, NGOs, and industry have contributed as much — if not more — than public servants to my policy artifacts so far.
- Ideally I’d love to set up shop in five Canadian cities for three days (Vancouver, Edmonton, Toronto, Montréal, and Halifax) each to workshop the Standard with stakeholders, provincial and municipal governments, and feds in regions offices. Maybe this could be recycled for different jurisdictions so that others don’t have to start from scratch.
So all-in-all, an awesome trip! I learned so much and met some great people who will undoubtedly provide pivotal input into my work going forward.
I was neither remunerated, nor did I pay, for either conference. Between the two organizing groups, I received gifts (a book, a magazine, and a bottle of wine) totalling approximately $50 in retail value as thanks for the time that I volunteered for these events.