Can the EU become another AI superpower?
economist.com
economist.com
One thing I've noticed in large EU corporations (where I and friends of mine worked as data analysts/"scientists"): upper management decides to invest in AI because they don't want to miss on this, so they create a new department ("AI/Robotics" or something cooler) and fill it to the brim with smart PhDs in mathematics and physics. They're all data scientists and ML engineers now, which means all the data mining, cleaning/preparation and labeling is going to be beneath them, it's not cool or impressive in their LinkedIn profile. They all want to work with the latest thing and each one has a different opinion and held to it pretty strongly. Nobody pays attention to product/project managers, they don't want to spend time creating PowerPoint presentations and dashboards to communicate and align with stakeholder. Discussing ethics in AI is a hippy silly thing. Then you end with something similar to what happened in my company: you create a bot to parse through CVs and decide which ones are better for any given job description. It took 4x more time than planned, and it's racist and discriminatory because it mirrors what the company did until now: hiring only certain kinds of people that studied specific degrees in specific universities that learned which keywords are good on a CV even if it means nothing. Nobody noticed or discussed this beforehand despite being so obvious because everyone is busy troubleshooting Keras or complaining about their GPU cluster.
You have absolutely hit the nail on the head there and this mirrors my observations of what's happening in the wider industry. Data science and AI are super cool viewed from the outside but the reality of the work day to day is that it is not fun. Getting a result and making a meaningful impact is very satisfying, but getting there requires careful, painstaking, meticulous work, getting the data and getting it into a format you can use is the vast majority of it. It is essential of course, but noone enjoys spending weeks (or months) decoding exactly what the fields are in this big weird CSV file you got from the mainframe and how exactly they marry up with the XML you got from this other system then doing that 100 times to mash up all your data sources, there's no documentation and the people who programmed it originally are long gone and get something that you can finally feed into your ML step. And then you come up with some recommendations which are immediately shot down because they are actually illegal and noone in the business can believe you even suggested it because that gets taught in Compliance 101 (I have really seen this happen).
Someone with the patience and the good attitude to do the data prep, and who has a bit of basic domain knowledge, armed with even the most rudimentary ML techniques will in any practical sense run rings around any rockstar researcher who just jumps in straight away with the AI. You would hope that PhDs who spend literally years doing research before writing up would understand this but it seems to be the first thing they forget!
You articulated so well something that I've been trying to say to my managers. Thank you for your post.
- There are some important industries that are very much interested in AI (for example the automotive industry). So serious private-sector money could be raised.
- Europe has the ability to pull of large-scale research projects and has the experience to do so. From LHC to the various institutions, like the Helmholtz-society or the DKFZ in germany (the EU also made some errors in preview large-scale research projects to learn from).
- The brains are there. I see many (very!) talented and dedicated students and researchers here and the research infrastructure (universities, non-university research agencies) is also established and quite diverse.
- I see that there's an understanding that entrepreneurship and non-traditional industries is an area in which the EU has been falling behind. I feel like it's improving.
I also don't think we're too late yet.
I see two main obstacles:
- lack of serious investment from public and private (this requires realising what a significant investment looks like). This is at the moment quite obvious if you follow the bmbf (german federal research agency). It seems like they don't realise how insignificant creating a few research groups is.
- no coherent strategy. Spreading everything thin without a thought where to reach critical mass is wasting time and energy. This is a problem especially in germany and, of course, the EU. We need a physical, European AI research hub with enough things like conferences and exchange to the other research-institute to get traction.
EDIT: What makes me really angry and frustrated (because in the end I am more or less powerless) is the complete waste of potential in germany. We have many great universities here with a lot of great faculty. But most of the universites are seriously underfunded and not really well-maintained. Some universities are so poorly maintained that their buildings are uninhabitable because of danger of collapsing. It's crazy, it's just laying waste. I think that we in germany wouldn't be in this situation if our universities could seriously compete and could enable all their potential.
But of course, this socialism will be much more successful than the old one. Just need more state money and state initiatives and state regulations, and so on.
So serious private-sector money could be raised.The EU needs serious reforms if it wants to be competitive in AI.
Academia in Europe is far behind academia in the US on average. And will be far behind China in a decade or so. European academics generally don't adapt to new technology. They have no reason to with fairly few links to industry, a funding that's based on personal relationships and politics rather than research, and an academic environment that doesn't really emphasize novel research. The same people, do the same things, for decades on end, with little to no progress or change. European academics don't put in much effort to get industry funding since your students are funded by the university much of the time.
Faculty hiring is very local and incestuous. German universities hire Germans, and only from a few places. English universities hire the British. French universities hire the French. etc. There are exceptions but it's rare.
The European PhD needs to be fixed. 3-4 years isn't nearly enough. It's hurting everyone. The moment someone gets productive they graduate. It's a total waste of time. They need to move to an American-Canadian style 5-6 year PhD. The fact that students are generally not funded by projects and researchers, but departments, also puts a big damper on people's motivation to hustle and publish.
Funding for startups in Europe is a disaster. It's really hard compared to the US and Canada, raising multiple rounds is harder, there's little infrastructure for doing it, and universities are little to no help. Rich people just don't have an appetite for risk, better to sit on your old money. This should be fixed by tax laws.
Pay is terrible in academia in Europe. Around half of what it is in the US and Canada at many ranks. When you can't live well, why would you stay in academia?
The tenure system is a disaster in many places in Europe and drives anyone good away to the UK, Canada, or the US. You have a lot of unpleasant steps where you aren't autonomous.
European research is also very closed. Europeans cite europeans, who go to european conferences, and do research with europeans. There are a lot of communities like this that are very closed and 2nd tier compared to international ones.
I could go on. Nothing will change any time soon unless governments take action to revamp the university system, university funding, and the tax code to encourage investment/risk. The next century won't be Europe's sadly.
The UK, Norway, to some extent Denmark, are much more open than say Italy, Germany, Spain, or France. Top-tier places in the UK are very open and international, mid-tier places aren't as diverse as mid-tier US or Canadian universities.
This is a cheaper and easier way to get industry insights than chasing companies for grants (which, of course, are not easy to come by here).
That is the third time I hear Merkel utter this disgusting sentiment:
“In the US, control over personal data is privatised to a large extent. In China the opposite is true: the state has mounted a takeover,” she said, adding that it is between these two poles that Europe will have to find its place.
It might not be that clear from above statement but a similar one left no doubt that some leaders have told her: For AI we need data and for data we need tracking/surveillance. So please look at ways to abolish privacy in the EU.
It's not lack of data which hinders business in the EU, it's overtaxing small businesses, cronyism, top-down-approaches, Google or something else which is hard to grasp. Lack of ambition? Lack of youth? Foreign espionage/sabotage? Negativity?
Angela, (if you read this, which I'm sure you do), you did a fantastic job protecting us from Cheneys torture doctrine, and much else afterwards. Erosion of privacy leads to erosion of societies. Don't mess it up in your last months.
>It's not lack of data which hinders business in the EU, it's overtaxing small businesses, cronyism, top-down-approaches, Google or something else which is hard to grasp. Lack of ambition? Lack of youth? Foreign espionage/sabotage? Negativity?
Different values.
Which is also what makes surveillance inevitable, as long as everyone is using Someone Else's Computer (the cloud).
Likewise, the health assistant needs your health data, but doesn't need to share it.
The data should be stored on hardware you control, and not shared with anyone else.
Likewise, I assume that every sufficiently important person is targeted with some form of surveillance by at least one organised criminal gang interested in either taking their money or corruption, both of which can be achieved by surveillance-enabled blackmail. “Sufficiently” is variable depending on much this tech costs, and it’s constantly getting cheaper.
But couldn't it also be that big Bay area companies aren't more American than they are international?
Europe is the opposite.
The EU is working hard to secure citizen rights though, but that doesn’t mean it isn’t working for ways that you can share your data. The EU just wants transparency and ownership to remain with its citizens.
That distrust is a reason why our government has been stable over the past 200 years, despite a series of technological, economic, geopolitical and cultural changes.
And, as an entirely separate point, because the information is too important — not just for exciting headlines about crime prevention (which is probably one of the last things it will fix in the general case rather than specific things like littering or speeding), but also for the mundane e.g. how to train an AI to do a production line job by watching people do production line work, or monitoring all substance use in real life to determine drug interactions.
I want a social, legal, and political system which takes this into account; one that tries to fight against it will fall to blackmailers, fully automatic payment frauds, and thieves cloning your door key from photographs.
Important point. Not only see I more hope for deep innovation in the manufacturing sector than in selling people the most targetted ads, this also has the potential to create much more equitable outcomes for everyone in the economy.
I don't really understand the concept of a 'AI superower' at all. Superpower at what, warfare? Concentration of wealth as AI returns flow to only a handful of people? Have Americans and the Chinese pondered whether there is some higher goal to the development of AI or just competition for competition's sake?
As a European(and German) citizen, I am much less concerned about taking the slow and steady route here. I have no interest in seeing Europe destroy its privacy or using AI to malevolent ends just to stay ahead in some fictional horse race.
People told us in the 80s that if we didn't move ahead with the service economy we'd be stuck in an archaic industrial society, and Thatcher was hailed as the reformer. I see parallels here to the AI debate. Now, where has this gotten the UK outside of London? For me, AI looks more and more like the hype around finance and services around that time. I'm okay with being a somewhat slow and bureaucratic grumpy German, if we get to be the guys who put advanced technologies into boring machines without much fanfare that's fine, people seem to keep buying them.
Yes and yes?
> Have Americans and the Chinese pondered whether there is some higher goal to the development of AI or just competition for competition's sake?
If they develop AI driven warfare techniques and we don't, they'll beat us. We must close the AI gap! (fwiw this terrifies me)
This whole debate reminds me also of the Japanese developments in the 80s. Everybody thought we'd be lost as they jump into the next superhuman age, funnily enough also involving grandiose computing projects.
I have the feeling this happens every generation and that the only thing that changes is the location.
In principle, in the sense of being able to give all of your citizens the ability to complete any task which previously only experts could achieve.
Google Übersetzer ist oft schrecklich und schlechhören(?) oft meine Worte, aber es ist immer noch deutlich besser als ich auf Deutsch, obwohl ich das Äquivalent einer guten Note an der High School habe.
> Have Americans and the Chinese pondered whether there is some higher goal to the development of AI or just competition for competition's sake?
Americans certainly have; its part of both utopian and distopian fiction. This fiction has been a guiding force with regard to what the AI looks and acts like in many cases, for example Alexa’s adverts were clearly trying to sell it as the ship’s computer in Star Trek.
I'm not concerned on the consumer side of things. We're all part of a global economy. If you want to buy AI services in Germany you can do that easily, just as I can use Google without a problem despite Germany not exactly being at the forefront of the tech. Having Google or Facebook physically located in your country, from a consumer standpoint, actually doesn't really matter at all.
Honestly this whole arrangement has always puzzled me from an American perspective. Americans sacrifice quite a lot in terms of equality, privacy and so forth to be at the forefront of this stuff, and we can just use it all the same because we're good customers. I'll be sad the day the US decides it doesn't want to be in charge!
Software (including A.I.) developed outside of your culture won’t reflect your culture. We already see this problem with regard to racism and sexism — A.I. which cannot detect black people at all, or hand-written healthcare software which cares more about insulin than periods — so for example, if you leave it to America and China, you won’t have any education software (with or without A.I.) which can cope with the difference between a Gymnasium, a Realschule, and a Hauptschule.
USA companies are also already having a lot of trouble with the cultural difference between the “easier to seek forgiveness than ask permission” attitude they’re used to in America as compared to anywhere which enforces rules because they exist for a reason. Google Street View in Germany, for example. I don’t know about Chinese companies, but I assume they also have cultural assumptions that won’t apply outside China.
The article doesn’t make a very clear distinction between academic and business AI, but I can’t see any inherent advantages for the EU in an academic perspective either-there are fields of AI that are under-represented in current consumer tech, but... I’m skeptical.
(I live in the EU and work in AI and ML-there is so much low-hanging fruit in terms of just making companies aware of what they can do that I seldom have deep enough engagements to step outside prepackaged approaches)
Contrary, whenever I open LinkedIn people seem to be naively super excited about Europe becoming an AI superpower (whatever that means). While I’m not against excitement, I can’t support it because it seems so detached from any real problem that we want to solve. And ultimately, that should be the driver for AI. Of course we can still engage in basic research, but we won’t become a AI superpower just because we want it.
The only problem for Europe so far, is that critical mass has not been achieved yet. Maybe it will eventually, hard to tell.
SAP, Nokia, ...
Search Engines, I don't think there is money to be made from this respecting the user privacy, we will see if DDG proves me wrong.
Nokia was destroyed by Microsoft, they forced Nokia to sell only Windows Phones and then MS bought them. But there are many electronics companies in EU for other domains, not sure if there is a need to compete with cheaper Chinese firms on phones.
I think it could be the culture, people may not be into get rich quick schemes and may want to work on more mingfull stuff then created a FB alternative.
The consumer side did suffer from Elop’s strategy (he also somewhat neglected the carrier network products), but the larger issue was that some carriers were so dependent on Nokia for mid-to-high-end handsets that when cheaper alternatives emerged they switched wholesale (I was at Vodafone at the time, and work at Microsoft now).
As to the rest, I don’t see Europe being a contender in chip manufacturing (even though ARM originated in our side of the Atlantic), and I don’t think patents would hold us back.
What does hold us back is the lack of significant manufacturing capabilities in those domains. There are a few electronics manufacturers and small fabs, but they cater to niches.
> BERLIN (Reuters) - The German government is taking steps to counter a surge in Chinese bids for stakes in German technology companies, including the creation of a billion-euro fund that could rescue such firms in financial trouble, a government source told Reuters.
Most succesful EU companies are 100+ years old due to taxes, bureaucracy and big company bullying. I find posts about the startup scene in most if not all EU countries laughable. Western Europe is too regulated and expensive (taxes) and Eastern Europe is too corrupt and politically unstable. Southern Europe is too hot and distracting. Maybe it has mafia as well. 20% VAT? Come on, that's outright theft. Hungary as 27% VAT. Portugal has 23% VAT. France has 20% but they also tax the crap out of companies and private citizens, especially if you're well off. Germany is over regulated, even the dogs barking hours is regulated. How do you expect tech startups to survive in this environment? Oh wait there's Ireland which gets lots of rain (people are cool working indoors) and has small taxes, except for VAT which is 23%. That one could work.
Just searching for "VW Steuern" reveals:
https://www.google.com/search?q=vw+steuern&ie=utf-8&oe=utf-8
"Volkswagen nutzt Luxemburg zum Steuernsparen"
"VW darf Milliarden-Bußgeld teilweise von Steuer absetzen"
"Der VW-Konzern nutzt ein fragwürdiges Steuersparmodell"
"Grüne wollen wissen: Wo zahlt Volkswagen wie viel Steuern"
"VW-Skandal: Steuern hinterzogen"
"Steueroasen: Wie deutsche Großkonzerne den Steuerzahler um ..."
This is only true if you're big enough. If you're a small business, you pay the inflated rate or go to prison.
But the thing about VAT is that it’s unimportant for most businesses. When selling B2B you just subtract all the VAT you paid from the VAT you owe. The only kind of business that is seriously affected by VAT is highly price-sensitive B2C, and technology startups usually aren’t that.
If you buy some computers for $100, do some work, and manage to sell your services for $1000, you get to reclaim $20 of VAT but you have to pay $200 of VAT, so a 20% VAT rate costs this hypothetical business 18% of its revenue, which is hardly "unimportant".
And a typical tech company would be taking in far more than 10x in revenue than what they are spending on VAT-able goods.
EDIT: But see below.
Can you name even one AI product that is sold directly to consumers? Most if not all is B2B and there VAT is not used as it is a tax on consumption levied at the consumer, not other businesses.
I thought pavlov was under the impression that the VAT you pay when buying automatically cancels out with the VAT you charge when selling, so the only people who VAT affects are end users.
Do also remember that all your competitors are bound by the same rules, so from a purely economical standpoint prices in a given country with VAT will just appear x% higher accross the board, and then the real question is more of disposable income.
To be perfectly honest I am no expert in VAT, but I do know that a country like Denmark with 25 % VAT is also ranked in the top 5 countries to do business in.
Also don't know what you mean by "As a business you stll have to buy stuff to recover VAT" but oh well maybe you can clarify.
In other words, "if you can't make at least a 20% profit on the value you add, you're not allowed to make a profit at all". This stifles innovation at the margin.
If you’ve ever done business in the EU, you know that prices are negotiated and quoted excluding VAT. The VAT rate literally is totally irrelevant to B2B pricing.
If you’re selling B2B from one EU country to another, you don’t even need to collect VAT. Another reason why its rate is irrelevant to EU startups.
Even there I'm not sure it matters, unlike corporate tax it's a local tax, so it's still an even playing field and it will have the same impact to everyone in the market.
Yes, in the US you can have something like 5%/10%. But it's not the main price determiner.
Also several company regulations don't exist for companies < 50 employees. Same as the US/Canada
I believe entrepreneurial and investment culture is a more important drive of innovation and startups than regulations. And yes, fewer people will give you an Angel investment in Europe for a murky business plan, while in the Valley you might get 100k for showing something in your laptop to an Angel over coffee.