Just to put this in perspective: Apple has about 30 times more cash unused than this German investment in AI. Plus this R&D budget is only "potentially" impactful.
The other day the French finance minister (Bruno Le Maire) said that the EU should form an empire. Implying that that is the only means the EU has to compete with the US and China going forward.
It's fascinating to watch all of this play out, the clash of ideas and approaches across the EU and Europe broadly, largely between EU-globalists and European nationalists that want greater sovereignty.
I have not read the whole strategy, just skimmed it.
I think one of the biggest problem of Germanies strategy (and the EU as a whole) is not trying to reach a critical mass (of researchers etc.) somewhere. You need a lot of ingredients for a center to suceed. Really good research, a lot of students, enterpreneus, funding for startups and strong industrial players.
I really liked the ELLIS-proposal. That should be the way to go. A concentrated effort with one big brand with a single hub in each (Ellis-)member state. Something that's recognizable.
12 centers are way too much for germany. I think 6 would be the absolute maximum, with 3-4 the optimal, realistic number (for example: Munich, Aachen, Karlsruhe, Tübingen). Then you could create some "applied ML" centers to distribute it further.
It's the same with the university excellence-strategy. 57 clusters are way too much. A tiered, more fluid system would be way better in my opinion, with a few "very excellent" universites, but also a way to build a profile and a way to establish key research-areas for the rest.
If you divide a sizable budget (and it's even not that sizable!) often enough, you end up with nothing. And germany always does that.
(Off-topic) I always see these big numbers from china, but I don't see any results in actual research yet. I have only come across a few papers from chinese universites and they were consistently not very good.
> it makes no sense for Germany to compete with France and the United Kingdom [...] A better plan would be to establish a new pan-European [...]
I am an outsider, but this sounds like a naive perspective. Competing locally has survival-of-the-fittest strategy that then can compete globally. However, taking public sector funds and treating these efforts as essentially government institutions that have an explicit goal of competing globally, you lose the organic growth and proof of value the market could otherwise help you decide.
The article makes a point that Silicon Valley is built with public funds, but to anyone really paying attention, it is clear it (and several other tech centers nationwide) is driven by private capital and grows with it. Same in many other states/cities that have thriving ecosystems. Imagine if policy shifted towards large amounts of public funds with the explicit goal of cross-state cooperation for global competition. That is just a government institution at that point, not grants.
I dub these policies "Envy-Based Funding" and I would be upset at my state/city using such large amounts of funds purely as a naive attempt to mimic SV.
Am I alone in feeling that there is a category error in comparing AI to HTTP and GPS? I have never seen it used effectively as a utility... badly and corrosively : yes; to generate value :no.
This AI hype is getting really annoying. It's sad how political decisions and allocation of budgets are basically just based on the trend. It's like kids wanting a toy just because some other kids have it. Instead of having long term visions, building solid technologies one by one, Europe just throw cash at the latest cool thing and of course it gives nothing. It's like the Human Brain Project, there was some hype in neuroscience so EU just threw cash at it & nothing came out of it. Now everybody forgot about it, nobody speaks of it & nobody cares because the trend has passed
In the UK, by law, mapping data access is controlled by the government.
Germany could declare, by law, that 1) AI training datasets and resulting networks are controlled by the government, which grants automatic permission to develop and use to all entities; 2) requires timely and no-appeal compliance with training data and network requests; 3) delaying publication of both for one year to keep the wolves at bay for startups.
They are unlikely to pass a law, and enforcement could never be truly certain or successful, but it would apply to Google if they transit German fiber. Germany could demand sanctions against the United States for refusal to adhere to their laws. It would be awesome.
Maybe someday.
EDIT: Another example, Tesla would be required to divulge to Germany both their training data (all driving data ever collected from a Tesla) and their AI networks (all versions) in order to continue transmitting data to German networks (or else a government with trade agreements, police, lawmakers, and a military is angry at them) with the understanding that it would be made public in one year (stagnate at your peril).
Now is the race for development of AI based war machines. Major powers will develop and perfect this technology and once thats done, United Nations is gonna be allowed to step in and it will do so by making development of AI based war machines illegal. What happened with nukes is being repeated. US, Russia, China, UK, Germany, France and probably India are gonna race for the development of these war machines before the development become illegal.
13 comments
[ 2.7 ms ] story [ 38.5 ms ] threadError message : Fehler beim Aufbau einer Datenbankverbindung
https://web.archive.org/web/20181116180324/https://ethicsofa...
It's fascinating to watch all of this play out, the clash of ideas and approaches across the EU and Europe broadly, largely between EU-globalists and European nationalists that want greater sovereignty.
I think one of the biggest problem of Germanies strategy (and the EU as a whole) is not trying to reach a critical mass (of researchers etc.) somewhere. You need a lot of ingredients for a center to suceed. Really good research, a lot of students, enterpreneus, funding for startups and strong industrial players.
I really liked the ELLIS-proposal. That should be the way to go. A concentrated effort with one big brand with a single hub in each (Ellis-)member state. Something that's recognizable.
12 centers are way too much for germany. I think 6 would be the absolute maximum, with 3-4 the optimal, realistic number (for example: Munich, Aachen, Karlsruhe, Tübingen). Then you could create some "applied ML" centers to distribute it further.
It's the same with the university excellence-strategy. 57 clusters are way too much. A tiered, more fluid system would be way better in my opinion, with a few "very excellent" universites, but also a way to build a profile and a way to establish key research-areas for the rest.
If you divide a sizable budget (and it's even not that sizable!) often enough, you end up with nothing. And germany always does that.
(Off-topic) I always see these big numbers from china, but I don't see any results in actual research yet. I have only come across a few papers from chinese universites and they were consistently not very good.
I am an outsider, but this sounds like a naive perspective. Competing locally has survival-of-the-fittest strategy that then can compete globally. However, taking public sector funds and treating these efforts as essentially government institutions that have an explicit goal of competing globally, you lose the organic growth and proof of value the market could otherwise help you decide.
The article makes a point that Silicon Valley is built with public funds, but to anyone really paying attention, it is clear it (and several other tech centers nationwide) is driven by private capital and grows with it. Same in many other states/cities that have thriving ecosystems. Imagine if policy shifted towards large amounts of public funds with the explicit goal of cross-state cooperation for global competition. That is just a government institution at that point, not grants.
I dub these policies "Envy-Based Funding" and I would be upset at my state/city using such large amounts of funds purely as a naive attempt to mimic SV.
Germany could declare, by law, that 1) AI training datasets and resulting networks are controlled by the government, which grants automatic permission to develop and use to all entities; 2) requires timely and no-appeal compliance with training data and network requests; 3) delaying publication of both for one year to keep the wolves at bay for startups.
They are unlikely to pass a law, and enforcement could never be truly certain or successful, but it would apply to Google if they transit German fiber. Germany could demand sanctions against the United States for refusal to adhere to their laws. It would be awesome.
Maybe someday.
EDIT: Another example, Tesla would be required to divulge to Germany both their training data (all driving data ever collected from a Tesla) and their AI networks (all versions) in order to continue transmitting data to German networks (or else a government with trade agreements, police, lawmakers, and a military is angry at them) with the understanding that it would be made public in one year (stagnate at your peril).