The brave new world of work is perhaps best understood with a hearing aid. A modern hearing aid can do much more than its ancestors, which only give new strength to weak hearing. You can now connect the devices to the sat nav in the car and have the route announced, you can pair them with your cell phone and have the conversation with a foreign business partner translated directly. You don’t even have to be hard of hearing. The small button in the ear can be linked to all kinds of devices. This will also happen more and more in companies, for example with a robot that talks to its skilled worker and alerts him when problems arise. This can be of great help.
But who is liable if something goes wrong? When the device overspeeds, an employee goes deaf because the hearing aid and the robot didn’t understand each other as they should? Does the manufacturer of the robot then have to take responsibility for this? The employer who uses it and may have neglected risks? The producer of the hearing aid? The TÜV, which tested the use of the machine but possibly did not take into account the connection with the hearing aid? The employee himself?
“AI will not be banned worldwide,” says the minister
Artificial intelligence (AI), for example in the form of the chatting robot, will more or less affect all professions. Developer competition is ongoing and progress is rapid. This raises new questions, such as liability in the case of hearing aids. But politics is lagging behind. Comprehensive AI regulation is dragging on. The EU Commission presented plans for this a good two years ago, but there is still no law. It looks like an uneven race between fast programmers and lazy legislators. At the end of March, more than 1,000 experts, including entrepreneur Elon Musk, had called for a temporary stop in AI development and for regulation.
“AI will change the everyday work of employees more and faster than ever before,” says Hubertus Heil. As Labor Minister, he is responsible for the impact of AI on Germany’s workforce. The Social Democrat does not believe in a ban. “AI will not be banned worldwide,” he says. “It is in our interest that Germany does not oversleep the development.” But how should politics, like the German civil service, keep up with all the AI developers who are receiving tens of billions from the capital market?
Heil climbed the wide stairs of his ministry on Wilhelmstrasse in Berlin to an oversized mezzanine floor. Nazi propaganda minister Joseph Goebbels once ruled here, later the first East German president Wilhelm Pieck moved in, and both dictatorships left their mark on architecture. Even today, this stairwell appears massive and as if it has fallen out of time. The visitor is all the more relieved when a few steps down behind worn doors graying bureaucrats are not crouching behind mountains of files, but rather an open-plan office spread out in a startup style: people grouped around new computer workstations, in the middle a table with a huge screen at the head end, high Videos of a tropical thicket and a surreally colorful cloudy sky run in an endless loop at the top of the walls.
A software scans data sources – even in Chinese
The brain workers of the work in the room “Denkfabrik Digitale Arbeitsgesellschaft”, a kind of early warning group of the Ministry of Labour. “We see a highly dynamic development, AI changes daily,” says Heil. You can see that with the AI text program Chat GPT: “We have to keep up politically here.” For Heil, his think tank is “our telescope” and “our observatory” for assessing the consequences of AI. And with the help of artificial intelligence.
Together with external service providers, the think tank people have developed the AI tool “Horizon Scanning”. The ministry has fed the software with subject areas such as “New technologies and their impact on work”, as they could appear with hearing aids, or “Effects of climate change on working society”.
In addition, the AI receives documents to read in, so to speak. It then creates search terms independently and scans 200 million data sources for developments in the world of work, new technologies or risks. Press reports, social media, but also patent applications and scientific publications are searched, even in Chinese. From this, the AI writes a summary, an editorial team from the ministry arranges and weights the reports and forwards them to the responsible ministry officials. “One advantage of the AI is that we are very quick with it,” says Heil.
In the end, people still decide
In this way, it is easier to identify important developments quickly, the think tank is convinced of that. And relieve the ministry employees who are not always aware of the trends on Twitter or the latest debates in specialist magazines. You will now get the hints from the AI. Horizon scanning has so far identified eight future fields. For example, “Technical improvements to the human body”, which also includes hearing aids, “Adaptation to climate change” with its consequences for work in the heat, or “Brain doping agents to increase performance”.
An AI that is supposed to warn of possible consequences of the AI - can that work out? What is ultimately important is still decided by people, says Heil. His declared goal is to make the world of work more humane through AI. “The AI knows no values,” he says.
For Heil, artificial intelligence is an instrument for better government: “We try to make political decisions based on scientific facts.” But politics is not a science. Politics means evaluating and making decisions based on existing knowledge, says Heil. “The more and the better the information is available, the higher the probability that we will make good quality decisions. Our AI helps us to filter out the most important information from the mass of information so that we can then make informed decisions.”
Heil says that in 2035 there will be no more jobs that have nothing to do with AI. He may then no longer be a minister, but he will still be too young to retire. The AI may be able to advise him at an early stage what trained politicians still have to do in this new working world.