The immigration use case is the most striking example of how the federal government is trying to automate the "boring" parts of the state. The Treasury Board actually updated its Directive on Automated Decision-Making recently to address these kinds of systems. While the AI can approve applications to speed up the backlog, the rules still require a human to sign off on any "high stakes" refusal. It is a massive shift from the 1990s, when Hansard records show MPs were constantly debating the backlog of paper files being mailed between regional offices. Now the bottleneck isn't the mail, it is ensuring the algorithms don't inherit the old manual biases.
Fantastic article, thank you! Especially the government’s own ChatGPT. Hopefully one day all of this can be stored in a sovereign Canadian cloud, unless one already exists.
The immigration use case is the most striking example of how the federal government is trying to automate the "boring" parts of the state. The Treasury Board actually updated its Directive on Automated Decision-Making recently to address these kinds of systems. While the AI can approve applications to speed up the backlog, the rules still require a human to sign off on any "high stakes" refusal. It is a massive shift from the 1990s, when Hansard records show MPs were constantly debating the backlog of paper files being mailed between regional offices. Now the bottleneck isn't the mail, it is ensuring the algorithms don't inherit the old manual biases.
I know Canada can’t be perfect, but this story warms me.
Thanks very much for alerting me/us to this database of AI initiatives, and for your efforts to highlight some interesting ones.
Fantastic article, thank you! Especially the government’s own ChatGPT. Hopefully one day all of this can be stored in a sovereign Canadian cloud, unless one already exists.
So encouraging. All these areas putting AI to good use. Many thanks, Leni, for such a good start to this day!