Many events tend to approach the topic of artificial intelligence (AI) from a consumer perspective. The 'Artificial Intelligence Forum' organized by WEKA Fachmedien, on the other hand, focused on industrial and embedded applications. The three media Computer&AUTOMATION, Elektronik and Elektronik automotive joined forces and organized an event with three parallel tracks.

The opening keynotes

In his opening keynote, Ralph Bucksch from IBM outlined the state of the art and began by clarifying the use of terms that are often confused: Cognitive computing, for example, is about making decisions interactively, taking into account detailed domain models, incorporating evident explanatory patterns as well as artificial intelligence and machine learning tools. Artificial intelligence, on the other hand, includes various techniques such as natural language processing, recourse to stored knowledge, weighing up and planning. Machine learning, on the other hand, is about statistical analysis for pattern recognition in order to make predictions based on collected data. AI is not always as easy to implement as is often promised. Bucksch admitted that IBM had overstepped the mark from a marketing perspective in the past. Many users thought it was enough to throw the data into a pool. But that is not enough, as Bucksch emphasized: "You need the rare data scientists and those who name the problem that needs to be solved."

In the second keynote, Volkmar Sterzing explained how Siemens approaches the topic of artificial intelligence in business. The main fields of application are in service (keyword: predictive maintenance), but the topics of operation in the life cycle, engineering and commissioning are also increasingly becoming fields of application for AI. To achieve its own AI goals, Siemens is already working with 200 data scientists and AI experts - spread across nine locations worldwide. Sterzing also presented an example in which Siemens has been active for a very long time - the topic of commodity price forecasts. Here, the company relies on recurrent neural networks. Thanks to AI software, the time of purchase can be selected in such a way that the raw materials can be acquired at the lowest possible price.

Dr. Bernd Kosch from Fujitsu presented a concrete practical example in the third keynote speech. He reported on how a manufacturer of turbine blades for wind turbines uses neural networks for quality assurance. As the turbine blades are only produced in small numbers, the challenge here was that only a few data sets were available for training. The method of 'transfer learning' using a convolutional network nevertheless leads to a high error detection rate and significant time and cost savings of around 80%.

AI in the factory

Following the opening keynote speeches, the event was divided into three parallel slots: 'AI in embedded systems', 'AI in automotive and telematics' and 'AI in the factory'.

In the latter, Prof. Claus Oetter, Deputy Managing Director of the VDMA Software and Digitalization Association, first gave an insight into the role of machine learning as a driver of innovation in mechanical engineering. Right at the beginning, he emphasized why AI is now making more and more sense: "More data has been generated in the last two years than in the entire history of mankind before that." And companies should use this data now and get started, he said, putting the onus on companies: "The AI train is moving faster and faster, companies can still jump on board, but at some point the others will be gone." His advice: "Just try it out!" There is plenty of open source software for this. "That's why I can only encourage every company: Try it out too!"

How companies can go about this became clear in the other presentations: Jörg Drees from Ifakt and Dr. Harald Bauer from Stuttgart University of Applied Sciences, for example, outlined the process model for applying machine learning in industrial practice - from recording requirements, data acquisition and preparation to modeling and implementation. A key message of the presentation: "A lot does not always help a lot." In one project, for example, effects occurred in individual data sets that contradicted the process description. In this case, reducing the data led to a significantly improved result.

Another common problem is that there is simply a lack of data from the past to train AI tools. Using a use case, Dr. Rainer Mümmler from Mathworks showed how companies can then proceed - with simulation using a digital twin.

Clever drives

Christoph Ranze, Managing Director of Lenze subsidiary Encoway, presented a specific AI solution at device level. This is not only used to predict the failure of drives or frequency inverters, but also to draw conclusions about the applications from the data - such as the conveyor belt.
Artificial intelligence also has an impact on the work of employees. "Our vision is to create intelligent assistance systems whose capabilities grow even faster than the complexity of automation increases," explained Dr. Alexander Maier from Fraunhofer IOSB-INA. This reduces the perceived complexity for the employee.

Markus Ahorner from Arhorner und Partner also had encouraging words in the final presentation of the slot: "80% of the way is done, the data is there due to the high level of automation". He continued: "We now have an architecture that allows us to collect data everywhere at the lowest level, we can create models at a medium level and we can create applications at the upper level where machines can basically take over 90% of routine human work via software agents, both in terms of optimization and plant monitoring." To round off the event, Dr. Joachim Schrey explored the question of who is liable for autonomous systems in his closing keynote. His sobering conclusion: "The legal profession is still at the very beginning when it comes to artificial intelligence" - in contrast to the technology. The next Artificial Intelligence Forum will take place in Stuttgart on May 19, 2019.