Volker Tresp received a Diploma degree from the University of Goettingen, Germany, in 1984 and the M.Sc. and PhD degrees from Yale University, New Haven, CT, in 1986 and 1989 respectively. Since 1989 he has been the head of various research teams in machine learning at Siemens, Research and Technology and became a Siemens Distinguished Research Scientist in 2018. He is Professor in Computer Science an LMU, filed more than 70 patent applications and was the inventor of the year of Siemens in 1996. We still have vivid memories of his Keynote Speech at SEMANTiCS 2016 which is why for the 15th-anniversary edition this year time has come to catch up. In this interview, Volker talks about real use cases, where he sees the most promising developments in Artificial Intelligence and the future of our kids.
Autonomous Systems and Machine Learning are two of your areas of expertise. SEMANTiCS 2019 General Chair Harald Sack recently said, that “Explainable AI is the is the key to progress and acceptance of AI.” Where do you see the biggest challenges ahead when it comes to enterprises employing Machine Learning, Autonomous Systems and AI?
The most important question is to come up with a business model that works for your company. I would claim that any company has successful AI projects. Machine Learning is a very strong and a very robust technology. But this does not mean that each company knows how to transfer this success into a business. Currently, no company, whose business is on the internet, can do without AI.
If it turns out that your company has an AI business, then you might have to drastically change the culture in your company. You can only be successful if you manage to hire the best talents and if you change the culture in your company so that it values, appreciates and listens to those talents.
Your recent SEMANTiCS Keynote was on “Learning with Memory Embeddings and its Application in the Digitalization of Healthcare”, so let us talk about healthcare: Hospitals generate huge amounts of data each day. What are patients’ benefits of structuring the huge amounts of information that is generated on a day-to-day basis? Do Knowledge Graphs play an important role here?
AI applications in image analysis have already been used for more than 5 years. Siemens Healthineers have been doing pioneering work and are working on many future use cases.
In general, I would suggest that the data situation in the medical area is as good or as bad as in any other enterprise. Mission critical data has high quality and the rest is typically “not well maintained.” The situation is different in clinical studies where personnel is paid to ensure high quality for study relevant data. This process can partially be automated but also involves a considerable amount of manual work.
I am completely convinced that knowledge graphs will become more important. As in the medical area, most domains consist of unstructured data (sensory data, images, …) and structured data (diagnosis, procedures, lab results, blood counts). Roughly speaking structured data is the domain of the knowledge graph and unstructured data is the domain of deep learning. A challenge is to have both talk to one another. That is what we are working on!
Where do you see the most promising developments in Artificial Intelligence?
Without any question, Machine Learning, and in particular Deep Learning (DL), was the reason for the recent breakthroughs in AI and it will continue to be the basis for the progress for the next years to come. ML builds on available labelled data, new algorithms, computational power and an incredible community. Machine Learning will interact more strongly with other approaches and we will see more hybrid solutions. For example, we and other research groups combine knowledge graphs with machine learning and deep learning. I am also convinced that the link to cognition will become stronger: one can learn a lot from the organizational structure of the brain. We as humans can manipulate symbols and this will also become more relevant for AI. But we have to go new ways and not warm up old ones. As I like to say: AI is not only Machine Learning, but without Machine Learning there is no AI. We are also working in my team on quantum machine learning; in particular, we are developing quantum machine learning algorithms for knowledge graph learning. Actual implementations of quantum machine learning are still a bit in the future.
Other fields, including Semantics and Knowledge Graphs, will increasingly contribute, but I am deeply convinced of the central role of Machine Learning.
What does the existence of ever-smarter algorithms mean for humans? How do we have to prepare our children for this future?
This is a difficult question. In the coming years, we need to build a strong academic basis for AI,in Europe, in particular in deep learning and machine learning, and we need to be open to new ideas that can contribute. The next generation is eager to learn about the new developments and we should bring Germany and Europe up to speed. In the long run, who knows? Maybe we will have a sizeable community of AI experts, but potentially the majority of the next generations might be involved in other activities like art, taking care of the environment, social activities and the like.
The annual SEMANTiCS conference is the meeting place for professionals who make semantic computing work, and understand its benefits and know its limitations. Every year, SEMANTiCS attracts information managers, IT-architects, software engineers, and researchers, from organisations ranging from NPOs, universities, public administrations to the largest companies in the world. http://www.semantics.cc