“Software can be very intelligent. But it cannot replace people.”
Rolf Heuer is confident that clever machines can help us – in research, industry and daily life. In an interview, the former director general of CERN and until April 2018 long-standing president of the German Physics Association (DPG) explains why we will still need intelligent people as well.
Artificial intelligence is a term that inspires hope and raises concern. What does it mean for you as a physicist?
Artificial intelligence used to mean simply intelligent software — neural networks that were increasingly good at identifying patterns thanks to growing volumes of data fed to them by people. Software was written in such a way that it could continuously advance itself with help from people. At first glance, this is still how it works today, but a second look reveals a difference: People are giving the software an immense quantity of data and writing it in a way that makes it is intelligent enough to comb through the data looking for patterns on its own and to select just the combination people need.
Do you have an example?
Let’s take the control system of the particle accelerator at CERN. It is equipped with millions of sensors to monitor the proper functioning of everything. Various sensors are called on simultaneously as patterns are recognized. The question is not: What is this particular sensor doing?, but rather: Which situation results? With intelligent software that combines the knowledge acquired from all sensors, we can quickly say: We’ll ignore that one sensor. But if the neighboring sensor displays the same discrepancy, the program has to be intelligent enough to determine: Now we have to do something.
So people no longer play a role?
They do. But these decisions need to be made in such short time that one cannot depend on people alone. People are too slow. But: The threshold for shutting off is set up by people, and they can change it. The machine cannot do that by itself.
The machine cannot or cannot yet?
A difficult question; one that the industry is exploring too and which involves a great deal of money as well as human safety. Whether for production lines, gas turbines or the infrastructure for trains: Sensors provide the basis for defining exactly the moment for shutting off. Up until now, people have been able to override the machine. It is certainly faster — but will it one day be smarter? Despite all the artificial intelligence, the value of human experience should not be disregarded. I cannot imagine that machines will one day override us. The same is true with autonomous driving: Software can be very intelligent, but it cannot replace people.
“Time to market is reduced and the products are better if you use smarter programs. But: Smart programs are written by smart people.”
Yet AI arouses fears, concerning the loss of jobs too.
There are always concerns like this when we have a new development; what happened with the first steam engine is happening now with artificial intelligence. It’s the responsibility of society as a whole, to determine: Does this represent a danger — for jobs, among other things? If so, we need to find out what we can do about it? And if not, that has to be explained. In my view, such decisions need to be based on proper, fact-based information. Science has to take a stand here, and this is what the DPG has always done.
So, what opportunities does AI open up for us?
What we are talking about is a cycle: In order to make advances, basic research needs ever more intelligent programs. Such new technologies naturally accelerate applied research and thereby drive industrial applications. With appropriately skilled people, this then advances the methodology, which in turn impacts basic research. This cycle is the most important thing we have. Because it consistently produces something that is disruptive, a true novelty — and it always happens at an unplanned location. The best example of this is the invention of the World Wide Web at CERN. There was absolutely no research money available for this. The better the programmed intelligence, the better the research can support it. The same is true for the industry: Time to market is reduced and the products are better if you use smarter programs. But: Smart programs are written by smart people.
So fears concerning machine superiority are unfounded?
Yes, if we stay informed and aware of our responsibility. There are many more dangerous areas: By using their cell phones, for instance, people are continuously disclosing information. That should be worrisome, but not AI per se. It’s always about the responsible use of technologies. It is our social obligation to ensure that we always stay ahead of the intelligent machines we build.
You don’t foster the hope that artificial intelligence will one day explain the world to us? After all, 95% of the energy and matter in the universe remains unexplained.
Artificial intelligence alone cannot explain the world. It is an instrument that man needs for making progress. As a researcher, I cannot accept any explanation, regardless how conclusive, if I can’t reproduce it experimentally. It first needs to be proven.
The Interview was conducted by Sandra Zistl
Subscribe to our Newsletter
Stay up to date at all times: everything you need to know about electrification, automation, and digitalization.