[peoplewindow]

Is that really true? Everyone in science speaks English and DeepMind, based in the UK, is in fact one of the world's top AI research hubs. Their papers are consistently excellent and of course DeepMind's workforce comes from around the world.

That's the European AI hub, right there.

The issue here isn't the language barrier. The issue is one that goes largely unstated in the ELLIS letter (it's a call for action, not really a plan). Which is this ... remind me what's wrong with working with the Americans again?

DeepMind was founded by a Brit - the guy who once worked on Theme Park and other much loved British games. He built up an absolutely top class research team out of nowhere, based purely on VC funding, then sold to Google not only because of money but because Google also built a world class AI research team out in California, and it made sense to join forces so they could work together on things like TPUs.

And why shouldn't the European AI hub have joined forces with the US AI hub? What benefit could there be from preserving institutional and funding barriers between Google and DeepMind? The letter sort of dances around the edges of this question, but ultimately the answer is purely politics.

It says things like

"This weakens Europe"

and

"we want the best basic research to be performed in Europe, to enable Europe to shape how machine learning and modern AI change the world"

and

"many European companies whose future business crucially depends on AI are not perceived as competitive"

These are all political goals, worse, they are not logical thinking at all - not a good look for a bunch of scientists.

AI research is entirely open. Having a new academic research hub in the EU-that-isn't-the-UK won't change the access those European countries have to the underlying research papers or techniques. They can use AI just the same regardless of where the research is being done. Maybe China would try to close up their researchers, but America and the UK certainly won't.

So if European companies are "perceived as uncompetitive" or are not "shaping the world" due to their poor use of AI (this is a dubious theory to begin with) then the solution is for more European companies to download TensorFlow and Torch and get cracking. It isn't for money to be poured into academics who will take it, write a few papers and then go straight into industry anyway.

[T-A]

> remind me what's wrong with working with the Americans again?

Make that American companies and the problem, from an academic perspective, is that they recruit away students and researchers from academia. That works for a few years, but longer term, the argument goes, it depletes academia to the point that new talent can no longer be trained... by academia.

At that point, the companies wishing to keep recruiting will create their own training facilities and/or step up collaboration with academic institutions. That's where the American part comes in: if the existing talent has all gone to the US and those making the decisions have been there all along, they are likely to focus their training efforts on the US.

On top of that, you have the well-publicized problems with data collection by foreign entities and local tax avoidance, which are indeed political.

[peoplewindow]

You make it sound like a zero sum game. If academics are being hired into industry, that frees up grant money to be spent on other areas where they aren't recruiting so aggressively. It doesn't actually 'deplete' academia except in the short run, unless the supply of students who want to do research becomes fully tapped out.

Also, people who are trained already can always go into academia or return to Europe. These movements aren't permanent by nature.

[T-A]

> If academics are being hired into industry, that frees up grant money to be spent on other areas

That's great for those other areas, but the open letter in question is about machine learning.

> It doesn't actually 'deplete' academia except in the short run, unless the supply of students who want to do research becomes fully tapped out.

... or the very limited supply of established researchers who can train them is tapped out. Also, "fully tapped out" is a squishy notion; you can always lower your recruitment standards, with obvious consequences.

But there is no need to get into hypotheticals here:

https://www.theguardian.com/science/2017/nov/02/big-tech-fir...

> people who are trained already can always go into academia or return to Europe

But why would they want to? They chose to move away for a reason, so something would have to change for them to change their mind. What would that be?

The authors of the open letter seem to think that a European AI hub could be that thing.

[gaius]

Academics consider anyone who has worked in industry “tainted”, such a person would never make it to tenure, even if they wanted to

Tenure is about the only carrot academia has to dangle

[DanielleMolloy]

I don't see that in computer science, or in my subarea of research. Major industry labs are highly respected contributors to science, and this has a long tradition with Xerox PARC and Bell labs, and now with Microsoft Research, Google Research and DeepMind (to name few among many).

[tnecniv]

That's not true. I knew a lot of professors at my undergrad, which is a well-known university, that worked in industrial research positions after their PhD before returning to academia.

[gaius]

These are all political goals, worse, they are not logical thinking at all - not a good look for a bunch of scientists

Indeed, it’s almost on the level of Make Europe Great Again