[mardifoufs]

Only $50k grants? You must have missed the AI startup wave here in Montreal. They didn't end up producing much, if anything, of value. But they sure knew how to scale those grants/"government investments" to a completely new level.

[alephxyz]

At least a few of them found a niche and are doing okay, but it was almost funny to see the number of AI startups 3-4 years ago whose business model was essentially "hope businesses pay us to apply ML libraries to their data".

[myth_drannon]

Yeah, that was such a sleazy move by these ML influences or whatever you call them. Have connections with government and the right degrees? Boom, now you taking photoshots with PM and syphoning hundreds of millions of dollars of public money. I think it's especially bad in QC where there is a lot of government support. There were a couple of startups in MTL that had absolutely nothing to show but a lot of PR and it was so so obvious.

[MonkeyMalarky]

I would've thought the whole Element AI debacle would've reined that in but I guess not? If the patron saint of DL can't keep a startup with 100M invested in it going, I don't even know what left, right, up and down are anymore.

[TMWNN]

>Only $50k grants? You must have missed the AI startup wave here in Montreal.

Indeed, I certainly remember the days when Montreal was said to be the next "AI superpower".

How much of that was driven by the media predicting after Trump's election in 2016 that hordes of tech companies and employees would flee north of the border? (And how much of that reporting, in turn, was driven by smug Trudeau government bureaucrats who actually believed what they told sympathetic/gullible reporters?)

[bmc7505]

Montreal still is a AI powerhouse. Most of the same people who were here ten years ago still are, and still working on the same research topics, just with less flashy PR.

https://youtube.com/watch?v=LVXkISpbq2k

[mardifoufs]

Yeah if anything I think the ecosystem is healthier. More established corporations have moved in, and the startups that pop up are usually more legitimate.

(Though I think MILA is kind of losing some of its lustre, both in terms of research output and quality. I guess it might just be that my friends there are mostly doing RL or cognitive research, which I both find to be pretty... unpractical? They give me a 1980s AI dead-end vibes, but that's obviously purely personal biais)

[bmc7505]

Interesting. Not disagreeing with you, but just out of curiosity, are there any other reasons why you believe research quality has declined at Mila specifically, outside the broader trend that progress in deep learning is starting to slow, and besides your opinion that RL/CogSci research are dead ends?