2025-10-20 AI Canada ==================== I’m reposting @mhoye@mastodon.social’s brilliant answers to a consultation on Canada’s “AI leadership”. (Source) The answers apply to all countries, needless to say. What efforts are needed to attract, develop and retain top AI talent across research, industry and the public sector? --------------------------------------------------------------------------------------------------------------------- Universal basic income. Excellent health care. Well-funded, accessible universities. Affordable housing. Clean, safe streets achieved by housing the homeless and sheltering and treating the addicted and mentally ill instead of punishing them. Thriving artistic communities in welcoming, multicultural cities. Progressive pricing for the fundamental necessities of life, like energy and shelter. Any computer can be operated remotely, over any distance, from any other computer. I'm doing that right now by filling out this form. You're doing it right now by reading what I've typed. Where the hardware is plugged in doesn't matter at all. If we want to foster and retain talented researchers, practitioners and public administrators in Canada, then we need to build communities where young people starting their careers can afford to live, where people growing their careers want to live, and where people in the prime of their careers are happy to live. What are the key barriers to AI adoption, and how can government and industry work together to accelerate responsible uptake? ----------------------------------------------------------------------------------------------------------------------------- Right now the key barrier to AI adoption is that the people pushing it the hardest are absolutely the least trustworthy people on the planet, with a small army of epistemologically-adrift FOMO-addled C-levels rolling up fast behind them, and it's being pushed into products and people's working processes with no regard to utility or productivity at all. Right now, the overarching purpose of large-model AI is not "improved productivity" or "reduced error rates", it is consuming enough computing resources, through whatever means necessary, that demand for computing stays high enough that the sector can keep their historic high-double-digit growth and returns going, and avoid turning into a commodified utility. But real-world uptake in most of these nonviable scenarios is close to zero and compute as a commodified uitility is where this is going. Accelerating responsible uptake means a measured approach, using consentfully obtained and carefully curated data to accomplish empirically sound and morally justifiable goals without freeing any person or organization using them from accountability for their actions. This means legal and regulatory frameworks with spine and teeth that understand what safe and consentful mean and impose meaningful costs for failing to meet those standards, and it means investing in historically-undervalued areas like criticism, curation and library sciences as foundational skillsets. When people depend on their tools – not just use but depend – then the liability avoidance that is absolutely endemic in computing is completely unacceptable. Car companies are simply not permitted to say, if it breaks you can keep both pieces. And it is the imposition of liability that has driven almost all of the modern innovations in automotive quality and safety. AI, and tech in general, needs those constrains in order to thrive. What needs to be put in place so Canada can grow globally competitive AI companies while retaining ownership, IP and economic sovereignty? ------------------------------------------------------------------------------------------------------------------------------------------ Again, I am begging you to understand two things: that any computer anywhere in the world can be operated remotely from any other computer anywhere in the world, and that landlords are a dead weight loss to society. If you want to foster Canadian innovation and champion Canadian companies, any approach to this that involves business incentives and lets the rentier class thrive is – inevitably, mathematically – a race to the bottom. If you want a new Renaissance, you have to build the new Vienna. If you want innovative new technologies to exist, you need to create communities where it is safe to fail. You need to create a society where the opportunity to, the ability to, and the tools to experiment are not the sole domain of already comfortable and already privileged. We know what those experiments look like already, they look like Mark Zuckerberg spending eighty billion dollars on the Metaverse and another eighty billion dollars on pretend chatbot friends and it doesn't matter at all and nobody cares. If you want to be the place that people go to build the future, you have to bet on people and build a place they'll believe is worth living in while they try to build that future. Invest in the arts, invest in the culture, invest in affordable housing and health care and tax the rentier class to fund those investments. How can Canada better connect AI research with commercialization to meet strategic business needs? -------------------------------------------------------------------------------------------------- This is a waste of time and effort. "Strategic business needs" are near-term problems whose contours and desired outcomes are known, where the remaining questions are about effective execution. Research is a long-term exploration whose results are always uncertain, but that sometimes pay off in wholly unexpected ways, even creating wholly unexpected markets. When the microwave was invented, not only was there no market for microwaves, but every single kitchen in the world already had a way to heat up water in it. It would have been the easiest thing in the world to say, what's the point of that, get back to work on stuff that matters. Focusing on the commercialization of research is the absolute latest part of the process. If you want real innovation, fund speculation. When something interesting comes of it, you'll get the commercialization for free in a few years. What lessons can we learn from countries that are successful at investment attraction in AI and tech, both from domestic sources and from foreign capital? ---------------------------------------------------------------------------------------------------------------------------------------------------------- Nothing. Right now, we can nothing from countries that are funding their own democratic, economic and ecological destruction by spending billion on circular investments into bubble markets without a real customer in sight, weaponizing automation against their working and minority classes and building vast data warehouses that will demand more electricity and water than their entire regions have available while driving the citizen's cost of both through the roof. We can learn nothing from them except how not to approach it and what not to do. How can Canada build public trust in AI technologies while addressing the risks they present? What are the most important things to do to build confidence? ----------------------------------------------------------------------------------------------------------------------------------------------------------- Consentfully obtained, accountably auditable training data, transparent (and open source) training techniques, and most importantly: liability. Most of our debates about freedom and safety are framed in terms of "freedom to", and the arguments are about the limitations on those freedoms. If you're luckier than me, those conversations might even involve "freedom from". But what if we lived in a society where that wasn't enough? Where government had a positive obligation to not just permit freedom to, but defend the possibility of? To mandate and enforce standards saying you must be free from the possibility of specific kinds of failure? We already have that for cars, of course. The full text of CRC c.1038, the Motor Vehicle Safety Regulations section of the Consolidated Regulations of Canada, runs a full megabyte of text. The Ontario Provincial Standards for Roads & Public Works is eight huge tomes and the first of them, General And Construction Specifications, is 1358 pages long. They detail – in excruciating detail – not just how the roads you'll be driving on must be built, illuminated, made safe and maintained, but how those standards will be verified, audited and enforced. Cars and roads built to these standards don't so much enable freedom of motion and freedom from harm as they delimit in exacting detail on what roads, at what speeds and under what circumstances people must be free from the possibility of specific kinds of harm, where their motion must be free from the possibility of specific kinds of restriction or risk. The amount of comparable legislation around AI is, I believe, zero. Canada could building public trust in AI by defining, not just what confidence could mean, but what a positive defence of that confidence and reliability must look like. By defining not just what trustworthiness means, but what specific possibilities of specific harm that trustworthiness means you are free from the possibility of. That might help. What skills are required for a modern, digital economy, and how can Canada best support their development and deployment in the workforce? ------------------------------------------------------------------------------------------------------------------------------------------ Respectfully, your real choices here are to either invest comprehensively in public education at the grade school, secondary and college/university/trades levels, or get some posters printed and pretend they'll accomplish anything. I'd rather you invest comprehensively in public education. (Shoutout to @grimalkina@mastodon.social on this one.) What can Canada do to ensure equitable access to AI literacy across regions, demographics and socioeconomic groups? ------------------------------------------------------------------------------------------------------------------- There's a famous metric in sociology about the effectiveness of social interventions in education called the "hamburger metric" – that is, if your social intervention can't be shown, decisively, to be more cost-effective at improving a child's academic outcomes than feeding them a hamburger, you should skip the intervention and feed them the hamburger. Canada can ensure equitable access to AI literacy across regions, demographics and socioeconomic groups by investing equally in access to AI literacy across regions, demographics and socioeconomic groups. Crazy, I know, but "every student has a clean, warm classroom in a clean warm community with clean warm food, shelter and clean running water" will do it. Which infrastructure gaps (compute, data, connectivity) are holding back AI innovation in Canada, and what is stopping Canadian firms from building sovereign infrastructure to address them? --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Let's be real here, you're not actually going to do anything about this country's insane telco trigopoly. How can we ensure equitable access to AI infrastructure across regions, sectors and users (researchers, start-ups, SMEs)? ------------------------------------------------------------------------------------------------------------------------- Another thing we can be real about is this: Canada already has two pieces of parallel infrastructure in place that provides equitable access to information across regions, sectors and users, available to every facet and lost geographic corner of Canadian society. They're the Canadian Broadcasting Corporation and Canada Post. The hardware is cheap and you can put it anywhere, so put it where we have places that already host hardware. Payment and delivery systems can be anywhere, and we've already got them everywhere. If you want equitable access to technology infrastructure and information, so people can start businesses that leverage AI tools from anywhere to anywhere, across Canada then start by leveraging our existing nationwide infrastructure, as a country, to do exactly that: The CBC should grow to become a citizens' web hosting service and Canada Post should become a bank. How can Canada strengthen cybersecurity and safeguard critical infrastructure, data and models in the age of AI? ---------------------------------------------------------------------------------------------------------------- Cybersecurity has approximately nothing to do with AI. If breach insurance companies declared today that your insurance was void if your software wasn't kept up to date, around 90% of cybersecurity breaches would end tomorrow. If they said your insurance was null and void if you didn't parameterize your SQL, injection attacks would end tomorrow. The recent Joint Cybersecurity Advisory from basically every intelligence agency in the Western world that came out last month warning about pervasive Chinese attacks and their success in achieving persistent access, had nothing at all to do with AI. If you go down the list of vulnerabilities being exploited there's nothing exotic or even interesting in it. Nothing, no zero days, no supply interdiction, no AI, no magic future technology from our ephemeral robot overloads, it's all "companies did not patch well-known CVEs on time", and those CVEs are all "programmer did not sanitise their inputs." This industry – and this country, and this world - are nowhere near the point where exotic knowledge or tools or magic are required to make huge progress on the cybersecurity front. This industry doesn't need MRI machines or mRNA vaccines or even penicillin. We need people to wash their hands. Let's start with that. Maybe if we're lucky we can even get people to wash their hands with soap. But that's where we are. People need to wash their damn hands. #Artificial_Intelligence