An interview about artificial intelligence and its limits with Prof. Dr.-Ing. Axel Sikora, Chairman of the embedded world Trade Fair Advisory Board.
Prof. Dr.-Ing. Axel Sikora is Scientific Director of the Institute for Reliable Embedded Systems and Communication Electronics (ivESK) at Offenburg University of Applied Sciences and Deputy Institute Director and Division Director "Software Solutions" at Hahn-Schickard Gesellschaft für Angewandte Forschung e.V. Since 2018, he has been Chairman of the Trade Fair Advisory Board and the Steering Board of the embedded world Exhibition & Conference.
Digitalization is called the 4th industrial revolution. How will it change our society?
This question does not only refer to the future, because digitalization has already changed our society significantly in recent years. New technologies and business processes, ever faster change, more and more cooperation between people and machines, internationalization and globalization. These processes will continue and intensify.
Technical innovations make life easier. Will digital technologies also make our lives better?
For me, a major advantage of digitalization is the increase in efficiency. Digitization enables a more economical use of natural resources, of time and of labor. While the first two aspects are certainly unreservedly positive, important social considerations must be taken into account when it comes to the "work" factor.
As an enabling technology, embedded systems are a key driver of digitalization. Where are they used and what development will they take?
The fascinating thing about embedded systems is that they are used everywhere, i.e. everywhere where sensing, controlling, regulating and acting takes place. Embedded systems are therefore an "enabling technology" in the best sense of the word. These applications will continue and intensify as the components become more cost-effective and energy-efficient.
Can you give an example from everyday life?
It is now almost difficult not to name an example from everyday life: It starts with white goods such as dishwashers or washing machines, continues with energy technology in photovoltaics, heating systems or electronic consumption meters, and of course extends to transport technology, especially automotive electronics, where a modern car has 60 to 70 embedded systems from the engine control and window regulator to the functions of autonomous driving.
Embedded systems form the basis for the Internet of Things. How far can networking go? (I.e. may everything that is possible also be implemented?) Can the network "tear" or become independent?
I believe that it is an almost Kantian basic rule that the networking of one may only go as far as the privacy of the other is not violated. But machines and factories also have a private sphere, which suppliers, customers, competitors or analysts must not, or cannot, violate. A second rule for me is that only secure systems should be networked. However, for various reasons this is not the case, especially in the area of embedded systems. This may be due to the fact that the systems are extremely cost-optimized, that they were put into operation many years ago when the topic of IT security was not yet taken so seriously, or that managing security for the embedded systems is difficult without a powerful user interface, i.e. if no password is entered because the system only has a few simple buttons.
Artificial intelligence is the "Next Big Thing". Where does research currently stand?
Research into "artificial intelligence" and "machine learning" is achieving impressive results, which so far have worked particularly well when the underlying data volumes are large. For example, methods for image recognition or social media can be trained with millions of data sets. In industrial applications, however, there are often only dozens or hundreds of training data available. Here, one of the challenges remains to achieve good results even with small amounts of data.
Are there any developments that worry you? Where do you see a need for action?
Oh, for all the enthusiasm for the topic of "digitization" and the "Internet of Things", there are a lot of important and critical tasks for the future. These range from the already mentioned security of systems and privacy, to the understanding and control of even complex autonomous systems, to the social and political impact of technological development, which has after all already begun in recent years. For me, a key challenge here is to counteract the "digital divide", i.e. the fact that many people do not participate in progress and are unable to do so. Unfortunately, this is true not only internationally for many regions of the world, but also for many people in our national societies. Here we must work to ensure that we can include them in the processes that are relevant to the future.
How and who regulates what an artificial intelligence is allowed to learn? Can and shouldn't an AI be taught ethical values in any case?
So far, no one is regulating this. In this context, however, I would like to mention two aspects. First, the legal system, and here both the legislative and judicial branches, lags miles behind in many of these technological areas, allowing leeway that is exploited by aggressive technology companies in particular. Secondly, it is also true here that technology in itself is not "good" or "evil" in the first place, but that it is the application that matters. And this ethical issue is very complex, but must always be kept in mind. The panel discussions at embedded world, but also at Net.Law.S, also in Nuremberg, are forums that drive the discussion forward.
Modern robots move like humans or dogs. They move and it appears that they can see and recognize objects like humans (embedded vision). What are the next development steps? What is still missing to become a real copy?
Indeed, machines are becoming more and more human-like. But as we know, in addition to intellectual intelligence, there is also emotional intelligence. We are seeing the first signs of this, for example in interactive service systems or nursing robots.
Can digitalisation be steered? What contribution do industry meetings like embedded world make to this?
One special feature of digitisation and the Internet of Things is that the development of solutions is as decentralised as they are in an endless number of companies and institutions. Central steering does not seem possible. Nonetheless, our state bodies are missing out on important room for manoeuvre here. This makes "family gatherings" like embedded world all the more important, where we not only discuss technical problems and present new approaches to solutions, but also take up strategic and ethical issues.
Thank you very much for the interesting interview!
Fotorecht: NürnbergMesse / Jan Scheutzow