A Story from Spring 2023 as Fall INDU Committee Approaches
Earlier this year, you may remember a small flurry of media activity related to a few letters, signed by concerned expert parties, about needing to slow down AI development/speed up regulation. There was one such letter in Canada in the Spring. Below is a bit of the backstory, as well as an ongoing effort to argue about the need to make space for other kinds of thinking on AI and regulation approaches. This example story may be helpful to keep in mind as Fall approaches, and with it, AIDA heading to INDU Committee.
Some Background and Context
On May 24th, a Canadian Press article, headlined: “AI pioneer Yoshua Bengio says regulation in Canada is too slow, warns of ‘existential’ threats”. In the piece, Bengio says he backs AIDA, and says the fact that it will come into effect in 2025 is too slow, urging immediate government action. He seems to take no issue with the idea of a government passing an undefined bill into law, then figuring it out after. He is, conversely, actively supportive of such a process.
If you read the piece, his position sounds reasonable. But the issue of what to do about AI continues to be the place we need to have more nuance. He is not wrong about the fact that there are issues with the use of AI. But it is highly likely that he is not the right person, nor appropriately situated, to direct an entire country on how to address them.
There is a persistent notion, a fallacy really at this point, that computer science experts are the ones to look to on how to address broad societal problems simply because they know the tech. Their AI expertise does not make them experts in democracy, it does not make them legislative experts, it does not make them access to justice experts, it does not make them policy experts.
There is also, in Canada in particular, major ignorance around the conflict we’ve got implicit in any of the AI discourse. ISED has been working since 2017 to support the growth of the domestic AI industry through something called the Pan-Canadian AI Strategy. It is now trying to also be in charge of laws to rein the industry in, so to speak. Also often missed is the point that Yoshua Bengio et al are located in a strategy that is funded by various levels of government to commercialize AI. It’s all the same team here. A closed circuit.
This conflict and lack of appropriate expertise for legislating should be patently obvious, in terms of the many limitations it creates. How pre-existing law could be applied is not adequately raised. How we didn’t regulate software — writ large — is not raised at all (there are specialized cases). How we would pull thinking about data and corporate power into the conversation is not raised. How we might consider the function and success of existing access to justice systems for existing tech harms is not raised.
This is all for good reason. A new AI law validates and makes legitimate the industry. It also globally harmonizes a range of AI applications. This deals a blow to the idea that we have arrived at the point where we need to have a much more expansive — and coordinated — look at technology and society and power arrangements of related infrastructure.
This is a necessary process to undertake, should we want to shift course from ever-more automation and efficiency culture. I think a lot of people want to shift course on that front.
One of the nuances that makes the AIDA approach really hard to talk about is not that it is actively majorly harmful, it’s that it is status quo entrenching and accelerating. Which is actively and majorly harmful. And this is happening at a time when we really need to re-examine the social norms and political economies around tech that are hardening as though inevitable.
The April Letter and Backstory
About a month prior to the May piece linked above, on April 18, 2023, The Globe and Mail published a piece titled: “Canadian AI experts issue letter in support of draft law aimed at curbing technology’s risks”. This piece featured a picture of Yoshua Bengio as the face of the story, and opened: “Canadian artificial intelligence experts and industry chief executives are urging Ottawa to pass legislation to regulate AI before the summer to deal with the potential harms, saying that any delay would be “out-of-sync” with the speed at which the technology is developing.”
“We ask our political representatives to strongly support AIDA,” reads the letter, which was co-ordinated by Mila, the machine-learning institute in Montreal. “Unless parties work collaboratively to move AIDA forward before the summer, we are looking at significant delays.”
“Mr. Bengio, the founder and scientific director at Mila, said in an interview there has been an “unexpected acceleration” with AI in the past few months owing to language-generating technology such as OpenAI’s GPT-4, along with text-to-image programs. “We really need public protection. These things are moving quickly, so we can’t wait another six months,” he said.” That was in April.
May. June. July. August.
“Mr. Bengio, who is also co-chair of Canada’s federal advisory panel on AI, met with Minister of Innovation, Science and Industry François-Philippe Champagne last week and urged him to speed up the timeline. (The minister was engaged but non-committal, he said.)”
The Story Behind the Letter — ISED + Mila
Below is the email that Mila sent out to some of the AI ecosystem asking them to sign the letter. I’ve redacted most people and company names because the point is more about how ISED operates, and how that shows up in the media, and in turn how public perception of the topic is created.
The Globe piece included mention of this email, but did not share it in full:
Date: Monday, April 17, 2023
Subject: Urgent initiative — Support of AIDA (AI & Data Act)
We have not had the opportunity to interact in the past. I am [redacted] at Mila.
I am reaching out following an emergency meeting of the Canadian Advisory Council on AI held by Minister Champagne at Mila Friday morning. The Minister is concerned and has solicited the support of its Council on Bill C-27. The Bill is being delayed by technical procedures by the opposition, and is at risk of being pushed back to 2024 if it is not adopted before the summer break.
You will find attached a letter that expresses support for AIDA. Our aim is to have it co-signed by 100 leaders of Canada’s AI ecosystem or industry, and to have it published in key newspapers across Canada on Tuesday. Bill C-27 will be discussed in Parliament again on Thursday, which explains the tight deadlines. I am soliciting your support, hoping we can add your name to the letter. I am genuinely concerned that if we do not rally now, it might take years before Canada legislates on AI, and we will have missed a critical global leadership opportunity.
My colleague Yoshua Bengio as well as Geoffrey Hinton will be signing [BW note: Hinton never signed) (the two Canadian Turing Award winners — the Nobel of computer science), alongside the CEOs of [redacted]. Many industry CEOs are signing [redacted] as well as over a dozen CEOs of AI startups. We have also confirmed the support of many professors including [redacted]
Please do not hesitate to suggest other names you think would be relevant and supportive. Signature is on an invitation-basis only, but we welcome additions. Out of transparency, we currently have <witheld> confirmed signatories.
I hope to hear back from you, and remain available for any question you may have.
I am not going to do a lot of analysis here. ISED is heading a strategy to commercialize AI. ISED is politically mobilizing members of the AI Advisory Council (which, to be clear, has the uptake of AI as its mandate).
This may be standard fare for industrial strategy and politicking. It all just continues to feel way way way too closed circuit for comfort.
As AIDA heads to committee in the Fall, it’s important to keep an eye on the stories being told, the way prior predictions on need for speed have (or haven’t) panned out, and the potential benefit of refusing AIDA and starting again outside of ISED.
Also a great time to read: Ana Brandusescu’s report: “Artificial intelligence policy and funding in Canada: Public investments, private interests.” Centre for Interdisciplinary Research on Montreal, McGill University.