Hacker News new | past | comments | ask | show | jobs | submit login
Google Brain Toronto (googleblog.com)
186 points by lebek on March 31, 2017 | hide | past | favorite | 66 comments



1. What is the Vector Institute? Is a foundation? Is it a not-for-profit corporation? Something else?

2. Who controls it? The government (which has put in most of the announced funding)? The University of Toronto? Google? A combination of all of those?

3. Who will own the work created by the Institute? How will it be licensed, and what rules will govern its publication?

These are questions that would be answered at a very early stage of an ordinary research partnership between a public university and a corporation, but I can find none of these details about Vector.


In another comment someone posted the link to Vector's website (which I couldn't find linked from any of the announcements, nor by searching on Google), which clarifies #1 and #2.

http://vectorinstitute.ai/

Vector is organized as a not-for-profit corporation. It has 12 directors including some academics, some big names from corporate Canada, two VCs, a few entrepreneurs, and a few people with close ties to the Ontario government.

The directors are:

Ed Clark - former bank CEO, Ontario government economic advisor

Scott Bonham - VC

Vivek Goel - senior administrator at U of T

Mary Jo Haddad - management consultant specializing in healthcare

Chaviva Hosek - U of T public policy professor, former Canadian government advisor

Jordan Jacobs - lawyer and founder of an AI company

Stephen Lake - engineer and founder

Nadir Mohamed - former Big 3 telecom CEO, associated with Ryerson University

Michael Serbinis - scientist and founder

Pearl Sullivan - Dean of Engineering at UWaterloo

Terrence Sullivan - healthcare policy professor at U of T, closely tied to the Ontario government

Shivon Zilis - VC


#3 bothers me a lot. It reminds me of the way owners of sports franchises use public tax-payer money to pay for stadiums, meanwhile keeping all the profits from the money made by their teams.

If the various governments in Canada are funding this research and Google is contributing such a minor amount of funding, who is the beneficiary of the research output? Does Google get first dibs on all the information? Does the research ultimately find its most valuable commercial use at Google? Who else stands to benefit from this research? Are any of these beneficiaries going to directly be Canadians? Or are Canadians merely going to be tertiary benefactors with the primary ones being the researchers working in Toronto taking salaries?


Google is one of 11 companies (the platinum level sponsors) that committed $5M to Vector. It's not a Google thing.

The researchers are Vector are free to pursue their own interests, and publish their work like any other academic. Their output is not owned by Google or any of the corporate sponsors.


Also, this is meant to retain talents in Canada rather than coming to USA to work for companies like Google. By working for Google in Canada? Doesn't having "Google Brain" in Canada imply people are working for Google but it's one based in Canada now?


Vector seems to be some kind of hybrid entity, presumably designed to soak up ML talent at the University that revived AI research while also soaking up taxpayer dollars for Google.

Added bonus of making Trump look bad by doing cool stuff in USA's backyard instead of the valley.


I am seeing a lot of investments and MASSIVE promised commitments by a lot of people/company. Is all this money going towards paying experts/ buying equipments ? Don't get me wrong, I like seeing all this money being poured in. It's just that i'm curious as to how this money is being used.

Tangentially, do you think any of this will trickle down to people who are trying to learn math/ML in the form of scholarships?


I agree. It has all been quite vague. I wonder if any will go towards AI based startups.


Well, I know ycombinator is doing it.[0]

https://blog.ycombinator.com/yc-ai/


Salaries.


Canada will benefit from the brain wall Trump has erected.


Those jobs would have been shipped to India eventually anyway. Global corporations don't care about silly things like stable economies. They care about the bottom line.

Edit: to those downvoting me, can you please explain why? Am I wrong in some way?


Just speculating, but positions in AI are usually filled by people with PhD's and this knowledge is so scarce in the job market, not low level IT pros who are competing in a race to the bottom


What keeps other countries like India from developing PhD's like that? It seems like they would have a strong incentive to do so.


Well in a sense they already do, but mostly at US universities.


I would imagine the quality of their university systems. I get the feeling that most top quality professors would not like to work for a university system run by the Indian government.


India doesn't have top universities and top research talent.


Do you have any source on India not having top research talent? From my understanding India has good amount of research and tech talent but it suffers a lot from brain drain to US universities. Especially students graduating from Indian institute of technology, national institute of technology + many more. In fact if you look into student (masters, phd) profiles of top US engineering schools like MIT, CMU, Stanford etc. most of the Indian students there have graduate degress from those Indian schools and they are in huge numbers. (who eventually end up at these companies or starting their own)

One of the main reasons for lack of research in India is lack of funding. Its nothing compared to US universities. But that does not imply it lacks talent. In fact primality test in polynomial time was founded by computer scientists at IIT - https://en.wikipedia.org/wiki/AKS_primality_test.


If the prophecy materializes into a exodus of brain from the Valley, I would expect very skilled and trained Indians to go back and, depending on the environment they find home, there would be no reason why their couldn't stabling top research and start ups in their homeland.


aren't IITs top univ.s?


Certainly some of them are for undergraduate education, but I'm not sure they're top research universities.


Maybe not now, but I assure you there many smart people trying to build the infrastructure and plenty of smart talent that could be utilized.


Ouch, sorry Waterloo.


UofT is one of the best research university in the world, at the PhD level.

Waterloo is best at producing the best engineers at the undergrad level.


Waterloo is best at bleeding great Canadian engineers to the USA


So does UofT - I myself and at least a dozen other 1T7 grads I know personally are headed south as soon as we graduate. This is a Canada-wide problem which won't be fixed unless salaries equalize.


Spot on. This new institute (and other govt efforts like it) all fail to answer this fundamental question - why would someone work at half the pay (or less) if they were a Canadian citizen? It is a chicken-and-egg problem .. salaries and labour pool quality go hand in hand.


Some of us enjoy living in a more equal society where people who are down on their luck still have the opportunity to afford healthcare and higher ed.


Equal society? Lets see you buy a detached house in Toronto if you are a first generation immigrant, and your parents are not loaded. If you do not own property today, I don't see a way to EVER afford it in Toronto. Our society is hell-bent on taxing income when we should really be taxing wealth. The latest budget is fairly against young people in favor of old people. But to each their own I guess.


>Equal society

Conveniently, you chopped the word "more" off of "more equal society". Kind of unfair to argue against a point the person didn't make.


Thanks for having my back :)


Being able to buy a bit faster doesn't make up for how badly poverty is treated in America. I would argue, in fact, that although it may be harder for a tech worker to break into ownership in Canada, your family can still take advantage of the countless social programs that aren't available to all hardworking people in America :)

And there's more to Canada than Toronto. Montreal has great rent prices and a fun-loving tech scene, and a very unique culture. Quebec has $5 a day daycare, free healthcare for all, and McGill university which is ranked as one of the best research institutions in the world despite having the cheapest tuition in North America (although prices for non-residents are quite a bit higher).


I never said Canadian society is perfectly equal, just more equal than US society. Sure property distribution is a massive issue but that's the case in any urban centre. There's a lot of progress left to be made but that doesn't mean we should just give up and head south.


Exactly. To invite people to stay here we need unicorn companies that young talent actually want to work for. I don't think this solves that problem as efficiently as possible. Would they be better giving that money to local startups in the space in hopes they become a unicorn and stay in Canada?


Till Canadian companies produce profits at US level, matching salaries seems like impossible. Same will be true for Europe also.


Because U of T engineers decide to stay in Toronto out of the goodness of their hearts...?

If an engineer stays in Canada it's certainly not because they're being adequately compensated.


Some of us would rather live in a country with better welfare even if we won't directly benefit from it. Compensation is good enough up here IMO.


There are three sponsor companies from Waterloo (granted, not the big ones) so not ruling out something in Waterloo


Well Pearl Sullivan is a director (Dean of Engineering at UWaterloo)


Right? This is an odd choice. Google just built their massive new campus in Waterloo, and University of Waterloo is one of the best engineer schools and computer science schools in the world, but they build this an hour away in a market that is much more expensive? An odd choice for sure.


I wasn't too surprised. If you look at the Vector Institute website (http://vectorinstitute.ai/#people), you'll see many additional well-known researchers involved, in addition to Geoff Hinton:

David Duvenaud, Sanja Fidler, Brendan Frey, Roger Grosse, Geoffrey Hinton, Jordan Jacobs, Tomi Poutanen, Daniel Roy, Raquel Urtasun, Richard Zemel.

there is a lot more momentum at UofT in this specific area of research, more than there is at Waterloo.


Hinton? Wow. After decades of going on about neural nets while most machine learning folks pursued other things, he's sure milking his 15 minutes of fame for all they're worth!


He has achieved "rock-star" status in academia - quite deservingly. I wonder how much more he will contribute directly to the field during the remainder of his career. I suppose once you reach his level focusing on these grander initiatives will help create a legacy.

Funny aside: I saw him walking up University Ave a few weeks ago. He was all bundled up in a coat and hood. It was cold out, but I also suspect he is becoming recognizable as a celebrity by some and prefers to keep a low profile. If I had recognized him earlier I probably would have tried to stop and chat.


Funnily enough, I was searching for an official Vector website (which isn't linked from either the Google or U of T announcements), tried Googling it and didn't find anything. So thanks for the link.


The University of Toronto is probably a lot better funded than Waterloo, too, has a massive student base, and has connections to other local universities that are also much larger than Waterloo like York and Ryerson.

This selection will certainly put a dent in Waterloo's reputation, though with any luck they'll be able to capitalize on it in some capacity due to their proximity.


Sure, but Toronto and Waterloo have different reputations when it comes to Engineering and Computer Science.

And if it really matters to you that Waterloo be involved, it's obviously worth pointing out that Pearl Sullivan (Dean of Waterloo Engineering) is on the Board of Directors for the Vector Institute anyway.


I think Waterloo would be an unknown school if it were not for their coop program.


> This selection will certainly put a dent in Waterloo's reputation, though with any luck they'll be able to capitalize on it in some capacity due to their proximity.

Highly doubt it. Waterloo is known for being mediocre/trash for CS research. People go to Waterloo to get into industry and not academics.


>Waterloo is known for being mediocre/trash for CS research

Are you out of your mind?

Do you know why Waterloo's School of Computer Science is named after David R. Cheriton?

Yeah, the same DRC who made a major exit to Cisco, seed funded Google and hold an emeritus professorship at Stanford.

A sizeable chunk of the people who published Spark have done their undergrad' or research masters at UW.

Waterloo is definitely top-notch at the undergrad', perhaps less impressive than schools with bigger endowment at the research level, but far from being "trash". Tons of Waterloo profs are on the steering committees of major CS conferences (VLDB most notably).

This comment is so puerile, it has to be written by some jealous and petty undergraduate student. Or so I hope.


The most recent Times Higher Ed world university rankings for Computer Science put Toronto at 17, Waterloo at 23. The QS rankings give Toronto 11, Waterloo 26. You can debate the merit of rankings, but there's simply no credible way to argue that a school that routinely appears in in the world's top 30 is "mediocre/trash".

Relatedly, Waterloo is in fact included in the Budget 2017 funding for Canada's overall AI initiative, of which this Vector stuff is a part:

"In this budget, the government announced the creation of the Pan-Canadian Artificial Intelligence Strategy, which will promote collaboration between AI hubs Toronto, Kitchener-Waterloo, Montreal, and Edmonton. It will be dedicating $125 million towards the launch of this strategy."


From the perspective of someone who's lived in both cities, it makes perfect sense. Toronto is a much much nicer place to live and it'll be way easier to attract talent to it. I wouldn't move back to Waterloo for any amount of money.


Why? I've never lived there but it seems one of the few places with tech jobs and houses that don't cost a million bucks.


Waterloo is great for backyards and parking. Toronto is great for urban living.

Take your pick :)


There's nothing there. It's a suburban wasteland.

Yes, you can buy a house there more cheaply than in Toronto. But is owning a pile of bricks worth living in a wasteland? Personal preference, as they say.


> Right? This is an odd choice.

Why? UofT / Hinton are doing the AI here...


Waterloo is good at churning out entry-level software engineers. UofT is much stronger for ML and research in general.


Geoff Hinton has deep roots with the rival University of Toronto and the surrounding ecosystem. His role and influence undoubtedly played a large part in this decision.


Few people want to live in Waterloo when they could live in Toronto instead.


It's part research center, part marketing expense...


"Posted by Geoffrey Hinton" it may be, but "authored by Geoffrey Hinton" if there's any truth to it at all must surely be true only in the most superficial and technical sense. With all the marketing speak in that piece it's completely alien to Hinton and very much in line with coming out of some corporate spin department.

Looking beyond the presentation though, I find this an interesting and important development for the field. I'm inclined to think there are plenty of valuable researchers who will work for the vector institute in Toronto who would not have gone to work for Google Brain in the SF Bay area.


Soo much money being thrown behind a methodology/philosophy that is potentially on its last legs....

So many bets being placed on gift horses hoping for a high return. So little efforts looking and/or thinking beyond a narrow box.

So much disruption potential outside of mainstream currents.

And to think.. All one needs is a computer and an idea.

What unfolds in the times ahead is going to be one for the history books.


Literally almost everything that happens, and everything in the past few decades of tech anyways, is "one for the history books"

History books contain a lot of inane details no one besides readers of history books are ever charged with knowing. (Then there's the important big-picture stuff and sub-arcs that we're supposed to remember lest we repeat them, but most people have already stopped caring by that point because of the inane details)


The phrasing "Canada's AI Moment" to me implies that this is temporary.


What [in technology] is not temporary?


It's a really strange phrase. Maybe meant to be "Movement"?


The influx is temporary, is probably what was meant by the phrase. When AI becomes as ubiquitous as, say, the internet, noone would call it AI.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: