Comment: AI – why ignoring and banning won't help
AI is annoying – but the typical German approach is wrong again. At the latest now, we have to be honest, says Martin Gerhard Loschwitz.
(Image: heise medien)
- Martin Gerhard Loschwitz
Perhaps you feel like I do and you just roll your eyes in annoyance whenever the keyword AI comes up. While chatbots like ChatGPT, Claude, Gemini & Co. don't actually have that much to do with artificial intelligence. Rather, they are very sophisticated statistics regarding language, which is why the term Large Language Model, or LLM for short, is more appropriate. However, marketing doesn't have the time or inclination for this differentiation. So it's AI up and down, all day long: AI on the smartphone, AI at work, AI at home, and so on. And far too often with a missionary zeal: Google, for example, now regularly serves up half-baked answers unasked as the first result for a simple search. If you scratch your head while using a computer with Microsoft Windows, Copilot immediately offers a list of dermatologists in the area, at least it feels that way. And anyone currently doomscrolling on Instagram & Co. is bombarded with ads for AI: marketing and sales? AI handles it now. Finding a partner? AI. Accounting? You guessed it...
Perhaps it is precisely the pervasiveness of the presentation that causes total annoyance in many places regarding the topic of AI. But that too is part of the truth: LLMs can already be used meaningfully and efficiently today. I myself have long avoided the common AI models until a question – more out of fun – entered into Claude impressively proved to me how well some everyday tasks can already be handled with LLMs today. And how much more efficiently.
Ignoring doesn't help, banning doesn't either
In line with this, there are increasing reports these days that German organizations are prohibiting their employees from using large LLMs or even technically preventing them from doing so. With sometimes adventurous justifications: For example, it is unclear who owns the created code. And of course, the sometimes noticeably overused argument of data protection cannot be missing, as if one were practically forced to share sensitive information with Gemini & Co. Such an approach certainly fits the picture in Germany, which is strongly averse to innovation. New things are first banned and then regulated to death. However, as usual, the approach does not solve problems, but rather creates new ones.
Because the fact is: Well-trained LLMs solve many everyday tasks better and faster than a human ever could in the same situation. One can think purely in monetary terms here and focus on the business administration aspect. Then it is hardly justifiable that a developer spends weeks on even reasonably simple code that Claude assembles in a few hours, and in comparable quality. Yes, Vibe Coding can be a problem. However, in the hands of good developers with real domain knowledge, Claude & Co. become a powerful tool. If companies and corporations refuse to use them, they unnecessarily tie a huge competitive disadvantage to themselves, in addition to the challenges that the German economy is already facing.
Which, incidentally, offers plenty of room for mischief. What would come of it if the balance sheet optimizers (pro-AI) of one consulting firm were locked in a room with the legal advisors (anti-AI) of another consulting firm to finally clarify the pros and cons of AI? Presumably, tickets for such an event could be marketed excellently.
AI solves problems – quickly
But one can also consider LLM from another perspective: In an increasingly complex world, AI can also be the tool whose use is simply necessary to find ever more complex solutions to ever more complicated problems. A current example impressively illustrates this: For months, the armed forces of Ukraine have been relying on cheaply produced drones networked by AI in the war of aggression against their country, which Russia has illegally started. This not only makes it possible to intercept Shahed drones, which cost many times more, with drones whose unit price is barely over 2000 US dollars – a return on investment that even bold business economists can hardly imagine. But quite practically, every projectile from Russia that does not hit Ukraine also does not murder people there. The principle works so well that the momentum is currently clearly with Ukraine on the front lines in many places. European defense politicians and military personnel have long been eyeing the solutions developed in Ukraine. This is a real success story that would simply not be conceivable in this form without AI.
Videos by heise
Discourse is urgently needed
Given the advantages, it is hardly to be expected that LLMs will disappear from the global scene anytime soon. Meeting them with categorical rejection, as can currently be well observed in Europe and especially in Germany, not only makes it impossible for oneself to benefit from the advantages of the technology. But it also prevents discourse on several questions that we as a society should have addressed yesterday rather than today.
It is already foreseeable that AI implies a clear-cutting of jobs. Tables have long been circulating that show which professional groups are likely to be affected first. Accountants, for example, are at the top of the hit list in various studies. And accountants in particular are an excellent example of the practical benefits of AI here. If I book a taxi in Berlin via Free Now, I have to divide the resulting receipt into three categories in accounting. The actual fare (7% VAT), the tip (0% VAT), and the service fee of the ride broker (19% VAT). If I want to correctly record a supplier in Lexware, I also have to correct it manually. This takes 30 seconds per receipt, during which I could just as well do something more meaningful. If I could instead throw the respective receipt at the AI so that it correctly splits and books it – what would be the added value of doing manual labor instead?
Typical STEM jobs are also obviously threatened by AI. And even venerable professions like lawyers are likely to be affected eventually. All of this harbors enough social explosive potential. Do you remember the debates a few years ago about what to do with unemployed long-haul truck drivers when autonomous trucks become market-ready? The AI discussion is essentially the same, but on steroids and with much greater social explosive power. And Germany, just like Europe, will not be an island of the supposedly blessed that can permanently withstand the storm of innovation. Instead, we must urgently begin the societal debate about how we want to cushion and handle the social impacts of AI, instead of passing the issue on unsolved to the younger generations. "After me, the flood" doesn't work. And hoping that the megatrend of AI will disappear again, just as little.
And on top of that, Europe – once again – has to answer the question of how it wants to position itself effectively in terms of AI. So as not to – once again – get run over by the technology giants from Silicon Valley. The Trump administration made it clear recently: Because they considered the new Claude models Fable 5 and Mythos 5 too powerful, they simply prohibited their export from the USA. Anthropic then had to shut down the models globally because even their own employees without a US passport were affected by the export restriction and it is impossible to determine which user of the software is a US citizen or not. If companies here in Germany use ChatGPT & Co. due to a lack of European alternatives, it is nothing less than another step towards complete digital dependence. And thus the exact opposite of the right and important digital sovereignty from Europe's perspective.
But how do you deal with AI?
AI is by no means the panacea that its creators like to advertise it as. But it can be used well and helpfully. Therefore, banning it is simply not enough. And the longer one stamps their foot here and postulates "I don't want to," the shorter the time becomes in which a model for dealing with the topic meaningfully can be conceived. The clock is ticking. For Europe. And for Germany.
(nen)
