AI - very practical and individual

AI is a megatrend and almost every IT provider is currently equipping its products with something AI-like. The Handelsblatt/Euroforum event succeeded here in an exemplary manner in separating the "wheat from the chaff".

Three medium-term insights crystallized at the two-day meeting: AI is already much further along than public perception believes; AI is a mass market in terms of frameworks and cloud computing; productive and verifiable solutions are emerging based on individual needs and expertise.

AI as a product is different from existing IT products, so it was not surprising that SAP was conspicuous by its absence, Siemens was very present, and perhaps the best and most sustained presentation came from SAP's existing customer Trumpf.

Trumpf showed a remote/predictive maintenance solution based on Microsoft Azure: In the sense of predictive maintenance and troubleshooting, the noises of a CNC machine are recorded using a smartphone and sent to a machine learning algorithm in the Azure cloud.

The neural network provides an analysis of the machine state. In addition to the innovative use of the AI framework in the Microsoft cloud, this project from Trumpf shows very clearly what can be meant by digital transformation based on current technologies: the optimization and adaptation of processes.

Traditionally, one might have connected to the CNC machine via a data line to query the status. However, this route carries the risk of "data misuse". A direct, physical connection between the machines and the service organization is neither in the interest of the user nor of Trumpf. The "Acoustics" interface mitigates the problem of possible misuse of CNC data.

The AI conference in Munich not only showed that users sometimes have the more innovative solutions than the providers such as SAP with the Leonardo framework, but also that the innovative providers themselves no longer use AI frameworks to establish solutions: They use AI to have AI solutions created.

The first neural networks for Machine/Deep Learning were still constructed by humans: Deep Feed Forward, Recurrent Neural Network, Long/Short Term Memory, Deep Convolutional Network and Restricted Boltzmann Machine.

Meanwhile, neural networks produce thousands of modifications of themselves and automatically sort out the inadequate ones. Visitors to the Handelsblatt/Euroforum conference also learned this.

AI is now available to everyone at relatively low cost via cloud computing and open source frameworks. Smaller AI concepts can be evaluated almost free of charge at AWS, Google and Microsoft.

As the example of SAP's existing customer Trumpf shows, there is no need for the proprietary framework SAP Leonardo. This made it obvious that the AI competence lies with the users, who, however, often and sustainably fall back on the support of providers such as Siemens or Bosch - as support in their own projects. However, ready-made AI solutions are very rarely purchased.