Canada's Innovation Underpants Gnomes Strike Again
Or, the federal government launches the Regional Artificial Intelligence Initiative and the AI Assist Program
In what I can only describe as a classic post on his One Thought to Start Your Day blog last year, Alex Usher compared Canada’s innovation programming to the Underpants Gnomes from South Park. He argued that at “a policy-structural level, Canadian governments have assumed that innovation policy = growth policy” and as a result Canadian innovation initiatives have a logic structure that amounts to:
Step 1: Give government cheques to people doing gee-whizzy things.
Step 2: ?
Step 3: Growth!
Unfortunately, it seems that a very similar logic structure is being deployed in two new programs, the Regional Artificial Intelligence Initiative (RAII) and the AI Assist Program.
Launched yesterday, the two programs are part of the $2.4 billion AI package from the spring budget, and which I wrote about last week. The RAII will invest $200 million over five years through Canada’s regional development agencies “to help bring new AI technologies to market and help accelerate AI adoption by SMEs and sectors across the country”. The AI Assist Program, meanwhile, will invest $100 million through IRAP “to help innovative Canadian SMEs that are building or actively incorporating generative AI and deep learning solutions into their core products and services.”
I rather fear that both fall into Usher’s Underpants Gnome trap. As Paul Wells wrote about another doomed federal technology adoption program, “We have a government that thinks announcements are results.” When it comes to these new programs we certainly have nice annouceables - seven ministers have quotes attached to the launch news release - but the details are fuzzy on how we get beyond the Underpants Gnome’s question mark and to actual results.
I have two main questions about the programs. First, I am unsure why there are two separate programs to begin with. One is about helping bring new AI technologies to market and supporting AI adoption by SMEs and sectors. The other is about helping SMEs develop and adapt AI and assisting them with “awareness, planning and execution to develop these technologies safely and ethically”.
Is there really enough of a difference between SMEs adopting AI vs adapting AI or between developing AI vs bringing AI to market to justify two programs delivered through eight separate organizations, with all the associated overhead? Given we have over 140 federal innovation support programs already (as Senator Colin Deacon has examined), why are two new funding streams required? How will companies know which is the most appropriate route for them or should they apply to both (with all the associated overhead of form filling for the company)?
My second question is, what are the intended goals for the programs? Against what is success being measured? These aren’t clear from the information publicly available. PrairiesCan has a list of expected outcomes for applicants for the RAII program that they must later report on, including the usual list number of highly qualified personnel jobs created in Canada, revenue growth, export sales growth, incremental private sector investment attracted, number of technologies to market, etc. But how do these roll up to the overall program? Will RAII have been a success if 100 HQP jobs are created across the country or 10,000? Do we care where in Canada those jobs are created? What if jobs are lost in SMEs because AI is being used to replace workers but that increases productivity? Is that a win? What if none of the technologies brought to market are successful at scale but instead just lead to the firms developing them getting bought out by foreign companies for their IP?
The underpinning rationale behind the programs appears to be that AI adoption and commercialization are clear positives and that widespread use across the economy should be encouraged. But this is from the same department that is seeking to bring in substantial new regulations of AI to encourage responsible practices. As Minister Champagne wrote to the Standing Committee on Industry and Technology last year about the proposed Artificial Intelligence and Data Act, “once AI technology already permeates our society, it will be difficult to change expectations or retrospectively address harms that have already occurred”. Yet while AIDA remains completely stalled in Parliament, with dwindling prospects to get passed, the same minister is launching $300 million in programming, “designed”, in his own words, “to serve as a catalyst for quicker AI adoption”.
Wait, what? Those two statements don’t seem fully in line with each other.
I’ve written before about the importance of policy coherence. As I say in that piece:
We need to acknowledge that we operate in a complex system and that decisions (or a lack of decisions) have downstream consequences.
We need to be transparent about those choices and have a real dialogue about what they involve. And we’re not doing that right now.
This seems another case where we lack the transparency and dialogue we need. This is a common theme. The Canada Digital Adoption Program was another example of technology-focused programming that lacked transparency in its design, failed, and was quietly shuttered rather than having any accountability or public lessons learned to inform new programming.
This thinking may have been done by ISED, and there may be strong theories of change and internal goals for these programs. There will presumably be evaluation frameworks in place. But we need these to flow out from departments to the public so the government can be held to account for the goals, the design, and the results of these programs. That is needed to move beyond the Underpants Gnomes approach to innovation in Canada.