When Professor Dr Antonio Krüger says “it is possible”, then the federal government is also allowed to pat himself on the back. Krüger works at the German Research Center for Artificial Intelligence (DFKI) as Chairman of the Board of Directors and is member of the AI Enquiry Commission of the German Bundestag. When Professor Dr Antonio Krüger says “it’s possible”, he has also laid down a learning curve himself. Because until then, it was not entirely clear to him that Germany was underperforming in Europe in terms of digitization in the healthcare sector. It was only after the Commission had looked at AI in the healthcare system that he realized that there was a backlog of work to be done.
In his function as technical-scientific director of the DFKI, he is responsible for an unemotional, hence an analytical assessment of what is happening in our country in terms of the degree of digitization. And he classifies this assessment with the above: “It is possible” and thus wants to say that the digital possibilities in the country are merely lying idle and merely need to be awakened. In other words, he attests to the federal government to be able to catch up on this diffuse backlog. “Crises are both water level indicators and motors,” he says. So patting the shoulder is justified.
Krüger appears as a connecting pearl in a chain called AI — The Rising of New Technologies. But the technologies are not that new. Nevertheless, there is a mysterious aura of connection, of togetherness and a certain sense of for-each-other. Everyone should pull together: Business, research, politics and society, says Krüger. And he is not alone in this opinion. Dr Thomas Thiele of Deutsche Bahn even says that AI is a political issue. After all, “we are also experiencing social upheaval through the use of artificial intelligence processes. Thomas Thiele is responsible for the AI strategy at Deutsche Bahn (DB).
If Krüger now says “it is possible” and is thus actually talking about something still to be done, why is there so much public talk about AI as if it already exists across the board? “AI and digitalization are often mixed up. The boundaries are not clearly defined. Sometimes people talk about AI, but what they mean is digitalization. Thiele can also confirm this and gives us an example from his company.
At Deutsche Bahn, AI determines the arrival of the requested train based on the state of traffic on the corresponding section. From 00.05.59 the train is then officially considered to be delayed and the colour in the app officially changes from green to AI-based red.
DB has been using classic machine learning methods, such as AI methods, to provide passenger information for around two years. A forecasting procedure is used to determine the punctuality of trains. “It is only visible to passengers via the green plus icon in the app. Otherwise, you won’t even notice this as AI. The norm-based forecast has now become an AI-based forecast procedure. And Thiele explains: “We must build up a framework to protect society from abuse. Just as our application is invisible, so are many others. Hence, we need protective conditions.
Simon Lochbrunner from MaibornWolff underlines this. Often you don’t even notice that software is being used that makes the decisions, see Netflix or Spotify. These platforms are now playing a major role in people’s taste formation. “We unknowingly put ourselves at the service of algorithms that provide us with recommendations, which we naturally click on. This is convenient, but it shapes us much more than we might realize. At some point, we save ourselves by not asking what we really want.”
Let us now look into the distance
Do crises give us a new perspective on AI? Probably true, “yes, I believe so,” says Krüger. He adds that the view has now broadened to include what digitization could do for society, and crises like this could also change the mindset of those who are clumsily faced with the issue. “So far, we have allowed ourselves the luxury of not taking it too seriously. What is happening now has positively surprised me. It shows that something is working in Germany. It is possible.” And he also conjures up a new togetherness.
Lochbrunner works at “MenschIT”. His company is committed to ensuring that, despite all the people involved in digitization projects, the focus must remain on people; IT must be interpreted in an anthropocentric way.
We ask again: Why does artificial intelligence of all things produce so much for each other?
What does it have that other IT movement did not have? For Krüger, it is the self-learning processes (machine learning) that are linked with AI. “In contrast to classical programming, here the programmer only provides the framework, but the learning itself is largely based on the data. AI correctly understood means an interdisciplinary view on things. “The data with which AI learns does not come from the programmer, but from the real world. And on this basis, the system then learns a certain behaviour.
Thiele also has a plausible answer: “It is the idea of automation that makes the difference. Because it is softened in learning processes. These processes cognitively develop a knowledge that leads to an action that then manipulates something in the real world. And these circumstances lead to the fact that a) exchange is important and b) the AI stands out from other methods.”
The old saying of computer scientists still seems to be valid “Garbage in, garbage out. If you put garbage data in, garbage decisions also come out. Does this take us to the ethical limit? “Yes, in some critical areas we should ask ourselves on what basis AI decisions such as granting loans or tumour treatment are made and consequently we would have to talk about discrimination. This is because it is already present in much of the data used for learning and then shapes the behaviour of the learning process.
Aha. And what happens next? Krüger: We need self-explanatory components that reveal why a machine proposes a decision, for example in the medical field. And he completes this view that we should remain alert. “At some point, the doctor becomes too lazy to question the whole thing every time. The system has often offered him good solutions. More often than bad ones.”
Krüger is not alone with this view. Lochbrunner admits. The results that an algorithm delivers must be explainable: No one can expect users to understand the highly complex algorithm. However, it must at least be possible to explain the pattern according to which it was trained and how it arrived at its decision. What sounds simple and banal has far-reaching consequences that go beyond pure programming. Krüger states: “We have to deal with the question of the extent to which algorithms are designed responsibly. Namely, exactly those algorithms that do tasks that were previously reserved for humans”.
Krüger is committed to a system design that supports human decisions. There should be a balance between the machine and the human being. If we manage to do this properly, it will be a huge gain for all of us. An important topic for Germany is human-robot collaboration in the industry. According to Krüger, AI systems unfold their splendour when decisions are fraught with uncertainty. They have learned to deal with uncertainty. Man and machine work together at the workbench. The robot brings repetitive abilities and above all its power, and the human brings fine motor skills and flexibility.
Bosch and Airbus are conducting research here together with the DFKI. The Innovative Retail Laboratory of Globus also aims to find out how AI and robotics can be used in the retail industry of the future. We are pleased that Professor Dr Antonio Krüger is advising the German government. The man knows his stuff. And in this respect, the government can pat itself on the back for having found such an expert.