A warning on pouring money into AI

Hey there, time traveller!
This article was published 15/04/2024 (718 days ago), so information in it may no longer be current.

Given how often government can hesitate to respond to new social or technological developments, it’s nice to see one be proactive for a change.

As the world grapples with what to do about artificial intelligence — how to control it, how to use it, or whether to use it at all — Justin Trudeau’s Liberal government has shown a willingness to meet the issue head-on, rather than waiting until the horse is out of the barn before starting to talk about policy. Since September last year, a proposed law, the Artificial Intelligence and Data Act, which would update privacy laws and place certain obligations on the use of AI, has been studied at committee.

Now, Ottawa is pledging heaps of cash toward building Canada’s capacity in artificial intelligence. Much of the $2.4 billion pledged will go into a fund which will give access to computing capabilities and technical infrastructure, according to the Canadian Press.

This is where it would be advisable, despite what we said earlier, not to go too fast in meeting the moment.

Artificial intelligence, in the mode currently being discussed, is a fairly novel technology and also a bit improperly named. AI models do not have “intelligence” per se — they can’t think for themselves, or think in the abstract, the way a person can. It can only be trained by being fed enormous amounts of data. After that, prompts given to it by a human user cause it to spit out what it believes to be the best answer. It has no ability to use its own discretion, or “know any better” and therefore if you wanted to, you could teach an artificial “intelligence” that up is down and left is right, or that an easy chair looks a lot like a 1964 Chevy Impala, and it would never be able to figure out the truth on its own.

This reality of how AI works is of particular chagrin to people in creative industries — artists, filmmakers, writers and so on. While early adopters relish the idea that they can create an illustration or even a short film with nothing but written prompts, artists rightly bristle that the “creation” the AI is outputting is really just a work of collage, a hodge-podge of other people’s work which has been fed into its system, clumsily smashed together into something “new.”

So it might be a little too soon to invest too many taxpayer dollars in “building capacity.” Trudeau says in the Canadian Press piece that $200 million will go to boosting adoption of AI in sectors such as agriculture, health care and clean technology.

It’s possible that with properly trained models, AI could indeed be helpful in some respects.

However, given the challenges these sectors already face, it’s easy to see how investing in AI as a solution to some industries’ woes could lead to disaster. There are real concerns which arise when considering the notion of handing, say, logistical tasks in running a hospital to an AI, however thoroughly data-trained. Since it cannot truly know who patient John Doe is in the same way a human can, can it truly be trusted to handle John Doe’s records correctly?

It’s possible we have only reached the tip of the iceberg of what AI can do, so there’s some sense in putting funds and energy into discovering its limitations. But it’s also possible that AI is nothing more than a high-tech carnival trick, a fad, a flash in the pan that can’t work and won’t work the way its proponents suggest.

So, while Trudeau and his government have been right to respond quickly to the dangers AI poses regarding our privacy, and possible criminal uses, it is perhaps best not to open the public wallet too widely before we really know if the bet will pay off.