Canada’s AI strategy misses the mark

Open this photo in gallery:

Fears of artificial intelligence tend to fall into two camps. On one hand, there is widespread anxiety that we are allowing AI to develop too fast, too far, beyond our ability to prevent or mitigate its potential harms.

These range from the threat to privacy, to the problems of telling what is real or human from AI-generated fakery, to the exploitation of children or vulnerable adults, to the more exotic (but not easily dismissible) fears of mass job loss or even human extinction.

The other types of fears are rooted in the idea that we are not moving nearly fast enough on AI, if by “we” you mean not humanity at large but our particular corner of it. These are the fears that other companies or countries are stealing a march on us, that we risk economic decline or military defeat or both if we do not press ahead as fast as humanly possible, or indeed inhumanly.

Where stands the federal government? To judge by its newly released strategy paper, “AI for All,” it falls decidedly into the latter camp. Of the strategy’s six “pillars,” only one is squarely addressed to the sorts of deep-seated fears that have contributed to the manifest public hostility to AI recorded in poll after poll.

Opinion: Canada’s AI strategy won’t build necessary trust

Even this, however, is framed less in terms of protecting the public than of softening it up for the coming dash to the singularity. “For Canada to thrive in the era of AI,” it declares, “Canadians need to trust in its promise.” So yes, yes, we’ll “protect Canadians, particularly children, against AI risks,” to the tune of some vague promises to update privacy legislation, pass online safety laws, and the like. There. Satisfied?

Then it’s on to the real business of the paper: hastening Canada’s rush to AI. Here again, the concerns fall into two camps. One, that Canadians are not learning to use AI fast enough, that adoption in this country lags far behind our rivals. And two, that Canada is not producing enough AI, that the technology, much of which was incubated in Canada, is being developed and commercialized elsewhere.

Of the two, the first is by far the most important. The historical evidence on this is clear: Prosperity does not depend on being the first to develop a new technology, or even to commercialize it, but on making widespread use of it, absorbing, diffusing and adapting new technologies from whatever source.

China was the first to develop gunpowder, the compass, paper and printing, but it was the European countries that put them to the most revolutionary effect. The U.S. invented the transistor, but Japan adapted it to mass production. And so on.

Seize the quantum future – or let the U.S. own it. Which way, Canada?

Or to bring it closer to home: Canada doesn’t have a computer-hardware industry to speak of, yet as users of computers, workers and firms across the economy have become enormously more productive.

The good news is that the government gets this, at least in part. The bad news is that none of the raft of new programs, policies and acronyms proposed in the paper are likely to have much impact. Devoutly as the Liberals seem to believe that nothing moves in this country unless the government subsidizes it, technology take-up is not as simple a matter as hiring more AI instructors. People and companies will figure out how to exploit AI’s potential as and when it becomes worth their while in their specific circumstances, and not before.

The worse news is that the government is still obsessed with building a homegrown Canadian AI industry, to the point of partially nationalizing (“take equity stakes in”) firms it considers promising. It speaks confidently of supporting “emerging national champions,” subsidizing startups, acting as a “strategic anchor customer” and other bits of industrial-policy gobbledygook.

The point here is not just that governments don’t know how to pick winners, or that nobody has any idea what AI is going to look like six months from now, let alone six years. It’s that the whole enterprise is misplaced. Contrary to much mythology, we don’t need frontier technologies or globally dominant firms to succeed.

Consider Austria: though it has few cutting-edge technologies or world-beating firms, it is among the richest countries in the world. Ditto Belgium, New Zealand, Iceland, and on and on. Why? Because their firms generally are very good at adopting other countries’ technologies. They do so not because the government pays them to, but because, as small trading nations, competition forces them to.

Digital sovereignty is all very well, if it means the ability to regulate the operations of technology companies on your soil. But if it means trying to compete with the U.S. and Chinese AI industries, it’s a fool’s errand.