Interesting. At first I thought this was going to be "Germany's R&D Institutes are each following a Gradient Descent algorithm, and ending up in local minima (as might be expected): the answer is to add Simulated Annealing". But you are calling for a bigger shake-up than that, if I understand correctly.
I do believe this is a funding issue rather than anything else. China and the US can produce these things far below cost. The US, in particular, has been dumping software on the world for decades, but is rarely challenged on its behaviour.
Europe has failed to protect itself, rather than develop itself (I would argue).
The connection between industry and research is extremely important. And it is very hard to image German research without industry funding it. If you look at German universities multiple of them are struggling, part of that is reduced funding by German industry.
The problem with recent AI advances is not that the German industry has no need for it, of course they do, there is enormous potential in engineering for AI applications (and even the slow rolling industry is realising that), especially LLMs. But the problem is that AI research, which german industry requires does not exist in Germany or even the EU. More so, research into LLMs is almost entirely done and provided by a few US companies. There is no one the industry *could* fund, even if they wanted to. The AI talent is all in the US and German research institutes have no expertise.
I think it is totally wrong to frame AI as something German industry is avoiding. But it is research which is not done in Germany and at the current pace will never be done in Germany or the EU.
Germany was, and still is, exceptionally strong in computer vision which is used extensively in their manufacturing systems. There has also been a great developments in agriculture applications.
I don't know why it doesn't get more awareness, and I don't know why it isn't commercialised more forcefully by European institutions, but I do know it's a mistake to focus on LLMs and "American tech" which spends such vast amounts of money promoting itself and convincing you there is no alternative.
„Germany played a major role in developing the software foundations of modern robotics, particularly through research institutions such as the German Research Centre for Artificial Intelligence and the Fraunhofer Society. Since the early 2000s, these institutes have pioneered robot perception, machine vision, sensor fusion, and human-robot collaboration. Their work enabled robots to interpret visual data, map environments, and safely interact with humans, so today’s warehouses can be automated, factories inspected, and robots do logistics and manufacturing for us. Only, not really in Germany.
While Germany advanced much of the underlying research, companies like NVIDIA built the core AI and simulation platforms used to train and deploy robots, while firms like Boston Dynamics and Amazon developed large-scale robotics systems and deployed them in real-world settings. Germany remains strong in industrial robotics hardware and automation engineering, but the scalable software stacks, AI ecosystems, and platform economics that now shape the robotics industry have largely been driven by US and Chinese technology companies.“
So I wasn't clear from you comment whether you believe that AI applications are now so closely tied to the foundation models themselves that there is no space for application-driven research. I'm uncertain on this myself.
I think right now there is enough space for innovation above the layer of foundation models, e.g. there are coding agents and many other products from companies which build upon the offerings of companies who train foundation models.
The question is whether calling these innovations „research“ makes sense, it definitely is not academic research. But actual academic research in those areas does not really seem to exist. Certainly the AI experts are not in academia.
German industry definitely has enough easily identifiable problems where llms can act as foundation for a solution, but the scientific complex which these companies have previously relied on does not offer these solutions. The radical shift, where the most important technology of the recent past has been almost entirely separate from academic research is extremely remarkable. (I am of course aware that much of the early theoretical work on neural networks has been academic)
Right now most of the impactful research is coming out of the frontier AI labs and a few startups. Whether that is an inherent property, I can not say for certain. But it definitely has significantly damaged one of the main pathways in which German industry embraces innovation, and I do not think that the blame can be put on the unwillingness of the corporations.
This turned out fantastic! Thanks for writing this. I had no idea about the German research ecosystem.
Interesting. At first I thought this was going to be "Germany's R&D Institutes are each following a Gradient Descent algorithm, and ending up in local minima (as might be expected): the answer is to add Simulated Annealing". But you are calling for a bigger shake-up than that, if I understand correctly.
I do believe this is a funding issue rather than anything else. China and the US can produce these things far below cost. The US, in particular, has been dumping software on the world for decades, but is rarely challenged on its behaviour.
Europe has failed to protect itself, rather than develop itself (I would argue).
The connection between industry and research is extremely important. And it is very hard to image German research without industry funding it. If you look at German universities multiple of them are struggling, part of that is reduced funding by German industry.
The problem with recent AI advances is not that the German industry has no need for it, of course they do, there is enormous potential in engineering for AI applications (and even the slow rolling industry is realising that), especially LLMs. But the problem is that AI research, which german industry requires does not exist in Germany or even the EU. More so, research into LLMs is almost entirely done and provided by a few US companies. There is no one the industry *could* fund, even if they wanted to. The AI talent is all in the US and German research institutes have no expertise.
I think it is totally wrong to frame AI as something German industry is avoiding. But it is research which is not done in Germany and at the current pace will never be done in Germany or the EU.
Germany was, and still is, exceptionally strong in computer vision which is used extensively in their manufacturing systems. There has also been a great developments in agriculture applications.
I don't know why it doesn't get more awareness, and I don't know why it isn't commercialised more forcefully by European institutions, but I do know it's a mistake to focus on LLMs and "American tech" which spends such vast amounts of money promoting itself and convincing you there is no alternative.
I wrote a bit about this here: https://seedsoftime.substack.com/p/german-carmakers-have-surrendered
„Germany played a major role in developing the software foundations of modern robotics, particularly through research institutions such as the German Research Centre for Artificial Intelligence and the Fraunhofer Society. Since the early 2000s, these institutes have pioneered robot perception, machine vision, sensor fusion, and human-robot collaboration. Their work enabled robots to interpret visual data, map environments, and safely interact with humans, so today’s warehouses can be automated, factories inspected, and robots do logistics and manufacturing for us. Only, not really in Germany.
While Germany advanced much of the underlying research, companies like NVIDIA built the core AI and simulation platforms used to train and deploy robots, while firms like Boston Dynamics and Amazon developed large-scale robotics systems and deployed them in real-world settings. Germany remains strong in industrial robotics hardware and automation engineering, but the scalable software stacks, AI ecosystems, and platform economics that now shape the robotics industry have largely been driven by US and Chinese technology companies.“
So I wasn't clear from you comment whether you believe that AI applications are now so closely tied to the foundation models themselves that there is no space for application-driven research. I'm uncertain on this myself.
I think right now there is enough space for innovation above the layer of foundation models, e.g. there are coding agents and many other products from companies which build upon the offerings of companies who train foundation models.
The question is whether calling these innovations „research“ makes sense, it definitely is not academic research. But actual academic research in those areas does not really seem to exist. Certainly the AI experts are not in academia.
German industry definitely has enough easily identifiable problems where llms can act as foundation for a solution, but the scientific complex which these companies have previously relied on does not offer these solutions. The radical shift, where the most important technology of the recent past has been almost entirely separate from academic research is extremely remarkable. (I am of course aware that much of the early theoretical work on neural networks has been academic)
Right now most of the impactful research is coming out of the frontier AI labs and a few startups. Whether that is an inherent property, I can not say for certain. But it definitely has significantly damaged one of the main pathways in which German industry embraces innovation, and I do not think that the blame can be put on the unwillingness of the corporations.