Hands-on workshops with a mobile team
The first KI-Studio opened in Munich in September 2023, and another was added in Stuttgart in February 2024. These hands-on brick-and-mortar workshops operated by Fraunhofer IAO are supplemented by the KI-Infomobil, a van that takes the project directly to businesses. The project team brings real-world demonstrators along and holds on-site workshops for employees to pinpoint where AI applications might make sense within their organizations and what steps are still needed to make these kinds of projects a success.
“Our goal is to put companies and their employees in the driver’s seat, taking away that sense of passive powerlessness. For them to be able to make sound decisions, theyneed to understand AI and the basics of how it works,” explains Dennis Klau, an applied AI expert at Fraunhofer IAO. To foster that kind of understanding and enable unbiased experimentation with the technology, the experts have equipped their KI-Infomobil with two kinds of demonstrators. The first visualizes concrete application scenarios, while the other type includes various technology demonstrators that illustrate how AI works. How does a language model work? What about an image generator? How does training data affect AI?
“We explain that AI isn’t a knowledge model, that a language model uses a probability distribution to determine its output, and that an image generator doesn’t just piece an image together out of individual elements of the training material. One of our interactive exhibits, for example, illustrates how AI uses pixel values to generate an image, step by step,” Klau explains. Another exhibit shows how training data affects AI by illustrating why a chatbot says what it does and how tweaking the training data changes how AI behaves and what it says.
A different demonstrator, this one from the application scenario category, focuses on document analysis: People can scan documents and find out which quote is the best match for a fictitious call for tenders. “The AI registers what it finds on the pages submitted, associates that information with the tender specifications, and then displays the result in a clear and easy-to-understand form. This is an excellent way to show how artificial intelligence really takes a lot of the burden of manual tasks like these off people’s shoulders and can also help make decisions more objective,” Pohl explains.
Fraunhofer Institute for Industrial Engineering IAO
