Artificial intelligence can take over human jobs

Digitization and AI : How artificial intelligence will change the world of work

The pop songs of the US singer Taryn Southern sound as pleasant as they are arbitrary and still or because of that inspire many fans: her hit "Break Free" has been played 1.8 million times on YouTube alone. What is special, however, is that the instrumentation was programmed by an artificial intelligence (AI). Another example of how rapidly the self-learning algorithms are being used in ever new fields was marveled at last week at Christie's British auction house: a portrait of a man in the style of Rembrandt or Vermeer's. However, it was not made by a Dutch master of the baroque, but by an algorithm. His signature can be found in the corner of the picture: "min G max D Ex + Ez". The work was auctioned for almost half a million dollars.

Are computers conquering the supposedly last human bastion of creativity? It's not that far yet. The painting software was fed 15,000 human templates and then tried to imitate them. Then an algorithm decided which of the creations were most similar to the human models.

Tagesspiegel Background Digitization & AI

Digital politics, regulation, artificial intelligence: The briefing on digitization & AI. For decision-makers and experts from business, politics, associations, science and NGOs.

Free test now!

"Strong" and "Weak" AI

Researchers differentiate between “weak” and “strong” AI. A weak AI can be trained to do very specific tasks better than a human. Here in particular, the progress has been immense. A “strong AI” that would be comparable to the general, intellectual abilities of humans, on the other hand, is not in sight.

Nevertheless, many experts compare the importance of AI for the coming era with the invention of the steam engine. The fourth industrial revolution will massively change the world of work. AI gives robots capabilities that were unimaginable a few years ago. According to a study by the World Economic Forum, machines should already be doing more hours of work than people by 2025.

Are there still enough jobs left for people?

The development affects classic office jobs at least as much as factory work. The list of professions at high risk include accountants, secretaries and tax advisors. Various studies predict massive job losses, especially in low-skilled industries. Others emphasize the emergence of new activities for which one must first of all be retrained. In many forecasts, the bottom line is even an increase in jobs that should also be less physically stressful and more fulfilling.

Not everyone believes in it. Microsoft founder Bill Gates calls for a robot tax. The idea: where robots will work instead of humans in the future, the state will lose income tax. Other entrepreneurs advocate an unconditional basic income, which is discussed around the world. In the event that algorithms and machines take on a dramatic number of tasks and people can no longer secure their lives with their wages.

An outlook on how artificial intelligence could change various areas of the world of work:

Area of ​​application 1: Industry and trade

Two robots have recently been buzzing through the long rows of shelves in Zalando's logistics center in Erfurt. They look like a rolling, modern version of machines at train stations. But instead of drinks or sweets, there are shoeboxes behind the glass cover. They use their cameras to scan the labels on the shelves. Once you have the product you are looking for, a gripper arm extends and stows the box inside. Zalando has been testing the robots for several months. If they meet expectations, significantly more of them should soon be working in the stores of the fashion retailer. "If we want to remain competitive, we have to use modern technology," says Carl-Friedrich zu Knyphausen, who is responsible for the further development of logistics at Zalando.

After all, competitor Amazon is much further ahead: More than 50,000 robots are used in the retail group's warehouses around the world. The group develops most of them itself: Six years ago, Amazon incorporated the robot manufacturer Kiva for 775 million dollars - and withdrew its mechanical helpers from the market and the competition.

This is probably one of the reasons why Zalando took a stake in the Munich company Magazino in the spring, whose Toru robot now has to withstand the first practical test. They still have to learn to precisely identify the shoeboxes, which are often colorful and differently patterned. "The robots have a self-learning algorithm and get better and better over time," says Knyphausen.

Pants cause problems for the robot

Online trading and the associated logistics have been growing for years. Tens of thousands of jobs have also been created for the unskilled. Today, ten years after it was founded, Zalando alone employs almost eight thousand people in logistics across Germany. Will it be over soon?

“I don't see right now that robots can take over from employees on a large scale,” says Knyphausen. The combination of humans and robots will also be indispensable in the years to come. History speaks for it: Especially in countries in which automation and the use of robots are particularly high, such as Germany and Japan, employment in the relevant industries has hardly fallen. Tens of thousands of robots have also been installed at Amazon since 2012. “At the same time, we created more than 300,000 additional full-time jobs,” explains the group.

Because with all the progress, the robots are still pretty limited. Just gripping different objects is an enormous challenge. The Toru models have been specially developed for cardboard boxes, but mechanical gripping arms have difficulties with soft, flexible objects such as pants or jackets. This significantly limits the area of ​​application. Because Zalando relies on the principle of chaotic warehousing. Clothes, shoes or T-shirts are spread across the warehouse. Bestsellers like blue jeans can be found several times in different places. This shortens the distances, the employees do not always have to go to the same shelves when looking for the orders.

Humans and intelligent machines work together

But the robots cannot collect pants and shirts on the way. Amazon is also familiar with the problem. For several years, the group even held its own competition and offered up to $ 250,000 in prize money to the developers of robots that can grab as many different objects as possible as quickly as possible. But now the technology giant from Seattle has apparently also stopped these efforts; this year the “picking challenge” was abolished in this form.

Humans and robots will work together more and more often. The electrical colleagues at Zalando, for example, are programmed to primarily fetch cardboard boxes from the upper and lower rows of shelves, which are more difficult for people to reach or have to bend over.

This type of collaborative robot is called “cobots”. A special model called Armar-6 is being developed at the Karlsruhe Institute of Technology (KIT). “The robot can recognize when a person needs help,” says Tamim Asfour, head of the humanoid robot research group at KIT. It was developed to support people with repairs and maintenance work.

For example, he helps a technician who is unscrewing a large panel from the ceiling to hold it and then carry it away together. But he can also hand him tools or other aids. It was not programmed accordingly for this, but learned its behavior by observing human technicians. “He is able to interpret situations based on people's actions and then offer support,” says Asfour.

Area of ​​application 2: Office and service

This year, 250 employees at Zalando also had to find out that AI not only takes on tasks in warehouses and factories that were previously done by humans. The group cut the jobs of marketing specialists. In the future, the tasks of advertisers will be carried out by algorithms. This development runs through many areas in the service sector: Chatbots automatically respond to customer inquiries, software automatically translates texts.

Jobs at banks or law firms, which a few years ago were the domain of well-trained academics, are also affected. So-called robo-advisors offer automated, computer-controlled asset management, and software can now read complicated contracts or even write lawsuits. The Berlin start-up Leverton offers AI technology that filters out crucial information such as notice periods or contract terms from long documents at the click of a mouse.

AI can do law too

"Software can now also read out certain clauses automatically," says Micha-Manuel Bues, who worked as managing director at Leverton for two years and founded his own start-up with Bryter in the spring. He describes his product as "Lego for lawyers" because they can model and automate decision-making processes with the Bryter software. A large Munich insurance company, for example, uses this for its 2000 internal guidelines - from rules to travel to gifts.

The Flightright portal also makes use of such technologies. The company checks for passengers after delays or cancellations what compensation they are entitled to. Flightright then demands the corresponding amount from the airline, if necessary also in court. The Potsdam company has already collected more than 150 million euros for its customers. The service is initially free of charge, but if it is successful, Flightright retains a commission of 20 to 30 percent.

Legaltech does not make lawyers unemployed

The model only works, however, because the cases are very similar and the process is highly automated: What a lawyer would have formulated in the past is now put together from text modules by an algorithm. In the meantime, the company has started to transfer its successful model to other areas. The Chevalier subsidiary was founded for this purpose. It is intended to help employers to get severance pay in the event of termination - especially in cases where comparatively smaller amounts are involved and those affected have so far shied away from hiring an expensive lawyer.

In the event of success, part of the severance payment will be retained as a commission. However, these so-called legaltech services usually do not make lawyers unemployed. They relieve them of simple tasks such as contract research or step in where lawyers were previously too expensive - and thus provide additional work on the other side.

Area of ​​application 3: care and medicine

When they see Paro, the old people cheer. The white, fluffy robotic seal moves its tail when it is petted and responds to voices with a squeak. Developed in Japan, “Paro” is now available in some German homes where people with dementia live and is not alone in this. There are robots that help lift patients out of bed and bring them to the toilet. We are working on models that can feed and wash.

"At the moment they are assistants from human supervisors," says Nicole Krämer, who as a social psychologist at the University of Duisburg-Essen researches interactions between humans and machines. “I don't think that artificial intelligence will take over old people's homes in the next ten years.” Caring for a person is too complex for that. The child-sized white robot "Pepper" is already rolling through hallways and explaining how an MRI examination works. However, he can only say what has been taught to him beforehand and does not notice when the patient is nervous or crying. Speaking, acting, feeling independently - from the psychologist's point of view, robots will not be able to do that in 50 years.

"Not everyone likes to be touched by people"

How do people react to the robots? “Some talk enthusiastically about their lives,” says Krämer. “Others are reluctant because they are just stupid machines.” One of their studies showed that when virtual computer characters smile, they are smiled back at. In another attempt, some participants were unable to turn off the test robot against its will. “It definitely has consequences when you equip machines with human behavior,” she says.

According to a survey by the digital association Bitkom, 57 percent of Germans believe that robots will help caregivers with heavy work in ten years. First of all, from the point of view of the ethicist and theologian Arne Manzeschke, who teaches at the Evangelical University of Nuremberg, it does not have to be a bad thing per se. "Many nurses are overburdened, some are annoyed," he says. "And not everyone likes to be touched by someone in intimate areas."

Nevertheless, Manzeschke urges caution. Robotic systems are currently being developed to help prepare meals in the home. Manzeschke is accompanying a project in which there was intensive discussion about whether robots should even serve the food. Technically, they could hold the fork in front of the patient's mouth. Then it would be difficult, for example, not to hit the teeth.

"But what happens when a person chokes and threatens to suffocate?" Asks Manzeschke. The robot could not help here and a person called for help would probably be too late. Who is responsible in the event of an accident? What is nursing work in essence when more and more tasks are delegated to robots? These questions should be discussed.

Will doctors soon obey the AI?

What is also being developed at the moment are virtual assistants like "Billie". He could ask an elderly woman who lives alone via a screen whether she has already drunk enough and taken her medication. But who can still see into the apartment with a camera like this? What will be done with the recordings and data?

Here too, critical questions arise. In addition, the degree of care is determined, among other things, by how independent someone is - and you are also considered independent if something can be done with aids. As a consequence, it is conceivable that more and more activities in the daily life of people in need of care will be supported by technical assistants and that visits by a caregiver will be saved. The older woman would be taken care of at home. “But from a human point of view she will always be more isolated,” says Manzeschke.

What AI still means in the health sector: Robots already assist during operations. In addition, AI can sift through vast amounts of patient data and compare symptoms and X-rays with thousands of others in a database. Everything faster and more extensively than a doctor can. Google's AI division “Deep Mind” has developed an intelligent algorithm that detects eye diseases more quickly. AI can also help identify tumors earlier. Although treatments can be improved as a result, Manzeschke sees problems: Doctors could increasingly see themselves forced to follow the recommendations of the AI ​​instead of forming their own judgment.

Now new: We give you 4 weeks of Tagesspiegel Plus! To home page