© istockphoto.com/salihkilic

A fundamental understanding of AI algorithms is essential

An interview with Professor Dr.-Ing. Christian Döbel, Steinbeis Entrepreneur at the Steinbeis Transfer Center for Integrated Systems and Digital Transformation (ISD) and Professor at the Gera-Eisenach University of Applied Sciences

In summer 1956, scientists at a conference at Dartmouth College in New Hampshire proposed that aspects of learning and other characteristics of human intelligence can be simulated by machines. The programmer John McCarthy suggested that this should be called “artificial intelligence” (AI). The world’s first AI program was written at around the same time, and was able to prove several dozen mathematical theorems. What is AI capable of in 2024, and how can businesses benefit from it? TRANSFER met with AI expert Professor Dr.-Ing. Christian Döbel to find answers to these questions.

Professor Döbel, you do a lot of work on production process automation. What role does AI play in this?

AI’s role is growing all the time, particularly in production process optimization, but also as a means of improving process reliability. For instance, we use AI to predict machine tool failure based on the current machine configuration. This allows us to adjust the configuration so that the time when the failure occurs is delayed for as long as possible.

By and large, however, the AI is always an add-on to the core processes, which are still programmed in the traditional way. The overall system must of course remain functional and verifiably meet all the relevant security requirements. We need to make sure this is the case in all our projects.

What are the specific benefits for SMEs?

At the end of the day, it’s about making value creation more efficient. We currently have a number of projects with SMEs, where we can get them up and running more quickly. For example, knowledge-based systems can help to address the skills shortage in a targeted manner by incorporating expert knowledge and using evolutionary algorithms to evolve autonomously.

These are very much pilot projects, although the tooling needs to be developed concurrently. But deriving other services from them can be useful for SMEs as a way of adding further value, for example enabling customers to review the resulting knowledge bases.

You can’t have a functioning AI without good-quality data. What do you see as the biggest challenges in this area?

Good-quality data is essential. We process all kinds of data. The first step is to translate it using parsers that we developed ourselves. Each dataset also gets a confidence score to ensure that

potentially “bad” data cannot be prioritized. The challenge is that we still have to process most of the data, even if it may not be very good-quality – otherwise we’d end up not processing 90 percent of it. So we are currently working with the businesses on the standards for data quality and the processing of both high-quality and poorer-quality data.

How do you think the growing use of AI will affect the role of people in businesses?

In all my projects with industry, I put a lot of emphasis on training the company staff who will eventually be using the systems. It’s important to achieve a deep, fundamental understanding of the AI algorithms, since you can’t just use them at the push of a button. You need to be able to think for yourself. As well as artificial neural networks, we mainly use learning classifier systems – and you don’t get that kind of know-how

everywhere. Our university is already teaching AI, but I also give voluntary AI classes in our region’s secondary schools. As well as that, I’m creating a scholarship program for schoolchildren so they can develop what I see as the missing core competencies that they will need for their future.

Contact

Prof. Dr.-Ing. Christian Döbel (interviewee)
Steinbeis Entrepreneur
Steinbeis Transfer Center for Integrated Systems and Digital Transformation (ISD) (Waltershausen)

227255-37