A man walks into a bar and announces that Germany will be a leading AI nation by 2030.
“Of course,” says the bartender. “Your table has been reserved for 2015.”
The government’s ambition is official. Germany intends to become an AI nation, with vastly expanded computing capacity. This is encouraging. Until recently, the national approach to artificial intelligence appeared to consist of regulating the artificial part and outsourcing the intelligence.
Now Berlin is grasping every available straw. Each straw gets a strategy, a steering committee and a photograph with the minister. Whether it floats is a question for the next legislative term.
Fortunately, Europe already possesses a business community exquisitely adapted to this environment: the professional consortium.
Its ideal product is a project that ends precisely when the next funding programme begins. Customers are temperamental. Grant calendars are dependable. Why endure the indignity of selling software when you can coordinate a transnational dialogue about the conditions under which software might eventually be sold?
Picture the funding aristocrat: fluent in Brussels English, permanently available for a panel, and able to turn three consultants and a dormant prototype into a European competence cluster. Yesterday, blockchain. Today, AI. Tomorrow, quantum resilience. The website changes; the catering supplier remains.
In this business model, success means continuity of funding. Commercial success is a delightful optional extra, like a hotel swimming pool.
This caricature has a factual foothold: the European Court of Auditors warned in its AI report that elaborate exploitation plans do not necessarily produce actual exploitation or commercialisation. Research deserves room to fail. But there is something wonderfully efficient about a system in which failure can become the justification for financing another attempt by substantially the same people.
The demonstrator demonstrates. The deliverable is delivered. The impact workshop establishes that impact requires further funding.
Nobody buys the product.
Fortunately, the product was the application.
The customer satisfaction survey is excellent, largely because the only customer is sitting on the advisory board.
Meanwhile, researchers who want to build things can compare salaries, computing resources and freedom to work across borders. Germany offers excellent science. It must also offer a reason to stay after graduation. “You will be joining a nationally significant stakeholder ecosystem” is a difficult proposition to put against a laboratory with available GPUs.
Then comes venture capital. Germany wants world champions, preferably without anyone experiencing an alarming amount of either wealth or failure. The founder should think globally, grow cautiously, disrupt responsibly and remain available during office hours. If the company succeeds abroad, a minister can always visit and describe it as evidence of German excellence.
The state could help by buying useful technology. Instead, imagine the procurement officer requesting ten years of references for a product invented last Tuesday. The startup qualifies for an innovation prize but not the contract. It receives a trophy, which unfortunately cannot meet payroll.
Next comes regulation. The ideal European AI would be powerful enough to transform civilisation yet incapable of doing anything that has not already been discussed. Even the iPhone becomes a suitable stage for this ambition: Europe will achieve technological sovereignty by holding increasingly sophisticated arguments about somebody else’s telephone.
Then somebody must build a data centre.
Enter our fictional yellow-bellied toad, occupying the proposed site with the confidence of an organism whose housing needs predate the coalition agreement. The minister promised global leadership. The toad requested a puddle. Only one of them submitted a realistic specification.
Protecting nature is reasonable. Announcing industrial expansion while leaving every practical conflict unresolved is stand-up comedy with a departmental budget.
Power presents another difficulty. Chips require electricity even when the minister is not visiting. Solar can contribute enormously, but a panel’s lunchtime output is not a complete plan for a November night. Storage, grids and dependable complementary supply still need financing and construction.
Perhaps Germany can train the first seasonally adjusted language model.
“Please regenerate your answer in June.”
A renewable electricity contract, meanwhile, is not a private extension cable to the sun. Someone still has to make the physical system work. This is deeply inconvenient for a political culture in which naming a transformation often constitutes its most advanced stage.
And surely the revolutionary fringe will welcome enormous facilities owned by technology companies, consuming electricity and processing data.
Naturally.
The 2024 arson attack on power infrastructure supplying Tesla, claimed by a far-left Vulkangruppe, offers grounds for including security in the budget. It does not make every protester a saboteur. It does suggest that “but ours is green capitalism” may fail to pacify people whose objection concerns the second word.
After all this, deployment remains. Somebody must clean the data, change the workflow and accept responsibility. Otherwise, AI merely gives the existing bureaucracy a chat window. Your application is still lost, but now its disappearance comes with a personalised explanation.
Germany has the brains and industrial experience to do much better. That is what makes the performance so irritating. Useful ambition would be measured in connected megawatts, retained researchers, paying customers and hours saved.
Instead, the danger is an AI nation whose most sophisticated learning system is the network of companies that has learned exactly how to win the next grant.
Back at the bar, the man asks whether his beer is coming.
“The pilot was extremely successful,” says the bartender.
“So where is it?”
“We’re applying for funding to explore scalability.”
Outside, the toad has water.
No comments yet