Artificial intelligence is the core element of solvatio. Roman Ernst has been with solvatio AG for 18 years. As a member of the Research Team, he deals with AI every day.
solvatio: Roman, artificial intelligence (AI) is one of the trend topics of our times. How did you come to be interested in AI?
I majored in biology, with a minor in computer science, and specialized in neurogenetics, behavioral biology and artificial intelligence. I started at solvatio’s back in 2001, applying AI methods for technical diagnostics. AI is a high-tech topic which fascinated me. As a member of our research team, I do detailed research into the application of state-of-the-art processes, and I drive technology within the company.
What has changed with regard to AI since you started at solvatio’s?
In my early days at solvatio’s, we used classic expert system methods such as case-based reasoning or heuristic classification. These still required a lot of manual configuration, which is why we replaced them over time with more powerful present-day methods, especially machine learning. Machine learning used to be around at that time, too, but we did not have the computer power and storage volumes to use it efficiently. And in most cases model training was impossible because the historical data records were insufficient in terms of availability and quality.
Taking stock: What can AI do today, and how do you see its role in the future?
Today, AI is already good at doing specific tasks, such as understanding language, translating, recognizing images, or – to give an example from our customers’ industry – diagnosing faults in Internet providers’ networks. In 10 to 15 years, AI may be able to solve more general tasks. Then, artificial intelligence might have the capability to recognize problems on its own, determine the necessary data and algorithms, and learn the solutions without requiring so much previous specification by human AI experts as today. On the way to reaching this goal, research currently focuses on Automated Machine Learning or AutoML. One of the goals of AutoML is the complete automation of the important step of data processing, which today is still a highly time-consuming job for experts.
How do you explain to outsiders what you’re doing?
(laughs) This is really quite difficult sometimes. Meanwhile, everybody has heard the term ‘AI’ – but there are many different ideas about what it actually is. Often, people think it is the same as robotics, and that my job is robot programming. Then I try to explain that while robots do in fact also use AI, I work in the field of AI software and counseling for other fields of application. But it is sometimes hard to make myself understood.
Which characteristics are most helpful in your everyday work?
Patience and perseverance! They helped my during my time at uni – and are helping me now. In my field, it is not easy to achieve sustainable solutions in an actual, active shop floor environment, because the parameters there keep changing all the time. For this reason, it is important to persevere in your work.
My job is indispensable because....
.... in the near future, practically every software in the market will incorporate some sort of AI component. At the moment, AI is top technology – it sets the course for the future.
What are the challenges when using AI projects?
One obstacle is trust: In the past hype phase, companies used to experiment and deploy AI in a wide range of applications – also in fields where basic preconditions were not met or where classical processes would have achieved better results. The success of some of the AI deployments was poor, which led to disillusionment. Word of this has spread across all industries and we now have to fight this loss of trust. At the same time, speech recognition, image recognition, and error recognition have made huge advances, which benefits businesses. For AI projects, the right assessment is crucial: What can the technology do, what prerequisites must be fulfilled and where are conventional methods better suited? This is where the challenge lies and this is also the point at which trust is regained through proper consulting and project support.
What are other sensible fields of application for AI?
AI does not only make sense in industry. In the medical field, for example, it helps with speech recognition and restores independence to handicapped persons. Coupled with intelligent robots, AI can also help paralyzed people to master their everyday lives. The areas of application are very diverse.
Pursuant to a Microsoft survey, Germany invests much less in AI than for example France and UK. How important is “AI Made in Germany” really?
Very important! If you forcefully push AI training and research, you also create the preconditions for its successful wide-range application. Germany should do more here. In Germany, companies of all sectors are looking for AI experts – and those are truly fascinating jobs which help to build our future.