Canada

Should've put the Canadian AI godfathers on the job

"Yo dawg, we heard Canadians had significant concerns about AI trust, transparency, privacy issues, and data sovereignty, so we fed your feedback without your permission into a bunch of commercial LLMs created by foreign tech oligarchs so we could figure out why you are so concerned about AI trust, transparency, data issues, and data sovereignty..."

JFC — the inmates really are running the fucking asylum..

Can someone that hasn't been brain-poisoned by the AI hype machine take this portfolio? No one is comfortable with a corporate kool-aid fountain running a government department.

I really don't see why Solomon is in that position apart from being Carney's buddy. It has never made sense to me as an appointment for such a critical role, and it makes less and less sense as time goes by.

integrated four large language models (LLMs)–Cohere Command A, OpenAI GPT-5 nano, Anthropic Claude Haiku, and Google Gemini Flash

Jesus christ, I wouldn't use those models to manage my grocery list. What was their budget for that, $1.50?

This is the actual summary from the surveys.

https://ised-isde.canada.ca/site/ised/en/public-consultations/engagements-canadas-next-ai-strategy-summary-inputs

As someone who reads these things semi regularly. This one is noticeably lower quality. It says the AI use was limited to certain aspects and things were manually reviewed but this thing almost feels entirely generated.

I'll file this under Carney picking another guy that can't even show basic competency at his role. Although that may very well be intended.

Goal to adopt LLMs and related tech seems driven almost solely by FOMO. Government, more than any other organization, should be applying healthy skepticism to claims of self-serving tech companies, most of whom have not even proven their tech’s worth, in the marketplace or anywhere else. Except maybe as a means of confounding the gullible and amusing the masses.

Every tool must prove it’s real worth helping people achieve a real and meaningful goal. Meanwhile, fanbois prove only that LLMs are good at cloning pre-existing products, at best remixing them, and introducing countless security vulnerabilities, all while losing any understanding of how any of it works, all in the interests of “moving faster”. But faster in which direction, nobody seems to know or care.