AI: Canada's National Strategy

[Debra Turnbull]

This has hit the airwaves today:
CBC article, June 4, 2026

Goals:

• protect Canadians and children
• free AI literacy training
• new jobs
• supercomputer

[Debra Turnbull]

... and here is one component to that strategy - data centres:
Supersized Data Centres

Okay folks => discuss!

[Chris Johnston]

@Debra-Turnbull

Google has beat them to the post-secondary market. AMII have a $5m grant from Google to provide AI Workforce Readiness support to all Canadian post-secondary institutions. It includes curriculum resource kits tailored to particular sectors (including nursing!) as well as an introductory online module, modest travel grants to Upper Bound, access to research etc.

While I think it's wonderful that all this money and support is being targeted at the post-secondary sector - I keep asking myself when is there going to be any support for the general public in terms of AI concepts - particularly in relation to health and healthcare?

I saw one mention of a National AI Literacy program through libraries - but no specific dollar amount attached unlike the headline items. Even then - no mention of any learning opportunities related to AI in healthcare for a public audience.

[Debra Turnbull]

@Chris-Johnston

I get a litttle leery when Big Business - Google, Amazon - get involved. Their angle is always => "how do we grow this to benefit us". There are $$$ big bucks involved here.

By growing the next generation of tech workers, they are preparing themselves for continued operations. Any inovational idea by these kids becomes the company's Intellectual Property (IP). That equates to profits. Then, when the markets tank and they need to trim their workforce, these tech workers are the first ones without jobs. History repeats itself.

I had to look up AMII. They're based in Edmonton (and Alberta wants to separate)... interesting. I realize that I'm wearing my "skeptics" hat but you are correct - where is the module for public literacy.

Wake up and smell coffee Canada!

[Debra Turnbull]

St. Mike's Expands across Canada

Healthcare AI needs a lot of data to work well. This project seeks to access data locked in silos across the country. The previous iteration - Gemini - showed that this is possible.

The idea uses "federated analytics" - it keeps the data local while the software runs on local infrastructure. Then the software "moves" to the next jurisdiction (or province). This is one way of working around Canada's interoperability problem.

So far, the model is only for research.

[Chris Johnston]

@Debra-Turnbull

I absolutely agree about big business and their hidden agenda - they never hand out money without expecting big returns.

But the fact is that the earlier Google investment kind of dwarfs the Canadian government's declared intention to invest - and both investments are being channeled through the same organization. Which begs the question, how can the public know for sure what their tax dollars paid for? Which is important, because if Google lays claims to discoveries and IP rights - we need to know whether that work was actually funded from our pockets.

[Chris Johnston]

@Debra-Turnbull

Fundamentally, I support interoperability, but I do wonder to what extent consent was sought from the 240,000 patients who became guinea pigs to the initial St Mike's project. I also wonder what kind of feedback (if any) was sought from them about benefits or adverse outcomes before the project was accelerated to expand.

What strikes me as odd about this particular initiative is that it talks about gathering anonymized data used for service improvement and research - that's not full interoperability because it doesn't allow for any individual patient to have multiple clinicians access a single record to avoid repetition or mistakes. From a patient perspective, that's the real win - everything else might be a nice to have, but as a lower priority. And if consent is being bypassed without the tangible promise of better individual patient care, then it feels very much like a bait and switch tactic, in which the patients lose out.

[Debra Turnbull]

@Chris-Johnston
Okay, valid.

See, here's the thing, data serves 3 functions:

• healthcare of the patient: locked into EMRs and HIS, you show up at their door = consent is implied.

• research, clinical trials: hospital researchers propose analysis projects - but first have to fly by REB; data tends to be de-identified. In the case of clinic trials - REB always; Informed Consent Form (ICF) always.

• and operational performance: Quality programmes. No consent required. Usually follows Plan, Do, Study, Act cycle; tends to be process oriented.

Now, when it comes to Big Data, I'm seeing data platforms springing up across the country. The purpose of these platforms escapes me - maybe research, maybe AI. The potential is there.

The Federated Analytics method is a cool way of solving the data silo problem. Instead of copying out the data and transporting to the processing site - processing happens locally. My question was - could this methodology be applied to solve the patient data fractionation problem.

I'm beginning to question if data for an individual travelling across the internet is a true solution (interoperability). Why not make the patient the data vector - where we are the true custodians of our data? We become "the binder".

[Chris Johnston]

@Debra-Turnbull

Yes, currently primary research (clinical trials etc.) is consent-driven and REB governed. However, secondary research - which aggregates existing data from various sources - often flies under the radar, not triggering REB involvement, and increasingly seeks to align itself with quality improvement/service strategy so that it escapes scrutiny for consent. But while this might be somewhat acceptable when tangible benefits for the population are clear, and profit is not the primary motive - there's often very little transparency about how much money is being generated or who benefits. So my personal stance is that there should always a requirement to seek patient consent for research and always the opportunity for individual patients to opt out of any research they don't find acceptable for any reason. Because if enough people opt-out of for-profit research, then it sends a clear message that we want the profits of research to be reinvested in healthcare and communities.

And while federated analysis can indeed avoid data sovereignty issues, and is in many ways more secure, I'm seeing a lot of corporate research saying that since the insights from federated analysis don't require specific handling of the data, again they don't need consent and there's no need for profit-sharing. But without patients, there would be no data from which to draw insights, and so again consent becomes a necessity. Since it allows us as patients to choose to withdraw access to our data from research initiatives that fail to recognize the need to share profits and benefits with the communities and populations whose data is mined to generate them. Now you might think that if a pharma researchers profits were spread thinly across a population, that might not make much difference to an individual patient. But if the beneficial ownership of the drug patent they develop is shared with health systems at the provincial or government level, then it would give leverage to produce that drug at an affordable level - rather than applying profit margins that run into the hundreds or thousands of percent.

It's feasible that a federated approach could potentially resolve the fragmentation of individual patient data - but the VITAL project doesn't appear to be tackling that as a priority. But in BC, where all the health system data (which includes hospitals, urgent care clinics etc.) is already co-hosted within a single mirrored data centre, it wouldn't be a huge stretch to start pulling in data from primary care, specialists and allied health professionals like PTs, OTs etc., under the same roof. That would make all of it more secure, and make the interoperability aim much easier to achieve. Interoperability doesn't have to be across the open internet, it can and should be achieved within closed, virtual private networks with appropriately deep layers of encryption and firewalls to keep the data and the networks secure.

I absolutely endorse the idea of patients in control of our own data, but I suspect that positioning us as custodians would simply allow other entities to disown their responsibilities. I saw a webinar on the NHS Front Door app recently, which claims to give patients more control. But when they spoke to patients, it appears to simply shift the bottleneck to an app that the patient can see - which they then need to keep explaining to various health providers - and if they forget it can go badly wrong.