A model to simulate how ideas move from one academic institution to another | Hacker News Reader

A model to simulate how ideas move from one academic institution to another

scientificamerican.com

52 pointsibhosszu8 comments

The actual scientific article: https://epjdatascience.springeropen.com/articles/10.1140/epj... (has a pdf link)

I am curious as to the esteem that the "Canadian mafia" institutions in deep learning were held in before they made their splash. I know UToronto is a premier Canadian university, but I always considered Waterloo the premier Canadian technical university, and neither has the esteem of the big four CS programs in the US. Which is to say that modern deep learning is a (happy) story of something coming out of left field. Maybe the relative isolation of UToronto in CS circles allowed more of a tinkerer mindset that enabled those researchers to survive the AI winter.

I don't think they're saying that great ideas can't spread from less prestigious institutions, I think they're saying that ideas from more prestigious institutions will spread further than similar ideas from less prestigious ones.

We intuitively know this though, don't we? In rhetoric we even have the fallacy known as the argument from authority: https://en.wikipedia.org/wiki/Argument_from_authority The idea being that because humans are prone to operating in hierarchies we give more weight to persons and institutions that are seen as more authoritative for whatever reason, one reason being their supposed prestige.

It's neat a study bears this out.

University of Toronto student here.

> I know UToronto is a premier Canadian university, but I always considered Waterloo the premier Canadian technical university

Waterloo probably beats us on the teaching practical stuff in undergrad, and definitely on the internship side of things. We definitely beat them on teaching theory (at least for the students who are interested), and research.

> Maybe the relative isolation of UToronto in CS circles allowed more of a tinkerer mindset that enabled those researchers to survive the AI winter.

I suspect isolation is the wrong culprit. I don't feel like we are especially isolated as a university for one. I think it's more the case that we just have a culture where it's normal for profs to be doing their own less than mainstream thing. Of course I may be misreading this, I'm a student here not a prof.

I don't think UToronto counts as "isolated" or "left field" by any means. It's definitely mainstream and current.

The success of the "Canadian mafia" might be better attributed to NSERC (the Canadian version of the NSF) cash-flowing research in the field for decades.

I did not know about NSERC funding deep learning, thanks for mentioning that.

Maybe "relatively isolated" was the wrong term. I meant to convey that deep learning advances didn't come from where you would expect, the MIT/Cal/Stanford places. Which is a good thing in that sometimes the usual suspects get stuck in a local minimum doing the same things and thinking the same way. And someone going off in a different direction, for whatever reason, can pop out of the local minimum and arrive at a better place.

Toronto has long been the premier Canadian University. I have no UT affiliation, but am a Canadian. It is not isolated in academic CS circles. It is not left field. It's the most prestigious university in Canada.

Consistently #1 in CS based on mainstream publications, for example, with a smaller faculty body than Waterloo:

http://csrankings.org/#/index?all&canada

Well of course. In society as well. It's called going viral.

Reply on news.ycombinator.com