An interdisciplinary project with the Bayreuth Industrial Museum aims to bring AI-supported games to life in exhibitions – technically feasible, but still with hurdles. What happens when language models such as ChatGPT or LLaMA not only answer questions, but actively change a game in the museum? Hof University of Applied Sciences is addressing this question in a research project that was developed together with the Bayreuth developer studio Emergo Entertainment and other partners – financed by the European Regional Development Fund (ERDF).
The focus is on an interactive game for visitors to an exhibition at the Bayreuth Industrial Museum. The vision: an exhibition in which the digital experience reacts to the visitors’ behavior in real time. A game that changes depending on the decisions made by the visitor – and is controlled by an AI. But how realistic is that? The Institute for Information Systems (iisys) at Hof University of Applied Sciences investigated this question. And the answer is: feasible in principle – but still a long way from a stable product:
It is basically possible for AI models to react dynamically to content and change it.
Jonas Kemnitzer, iisys
Long-term partnership with patience
The collaboration with Emergo Entertainment began several years ago – long before generative AI became a trend. “At that time, the idea was still vague, it was initially just a letter of interest to apply for project funding,” explains Paul Redetzky, Managing Director of Emergo. The actual research question – namely whether the so-called gamestate within an exhibition can be changed individually for each visitor, i.e. a direct interaction and reaction between game and visitor – only emerged later. a first practical attempt followed in 2024: a temporary exhibition was realized together with the Bayreuth Industrial Museum. Three foundations provided financial support for the project, which was designed to be both playful and educational. The AI was used for the first time, albeit in an early version (GPT-3.5). “The game was solid, but the AI could not yet meet the didactic requirements,” says Redetzky. The controls seemed too linear and the desired dynamic – the feeling of moving freely through the exhibition – was somewhat lost.
Technology advances
In the meantime, the technology has progressed: experiments were carried out with newer AI models such as LLaMA as part of the ERDF project. However, the challenge remains the same: how can an AI be integrated into an exhibition that does not yet have a permanent home? This is because the Industrial Museum in Bayreuth has so far existed primarily as a concept – changing exhibitions at different locations, but no permanent site. A permanent museum is only planned for the coming years. The new technology could be used for new or revised versions of the exhibition.
Outlook: Much learned, much to do
Even though the full potential of the technology is yet to unfold, those involved are already optimistic about the results so far. “We have learned a lot, which we can now incorporate into the application in a targeted manner,” says Paul Redetzky from Emergo Entertainment. The knowledge gained will flow directly into the further development of future games – not only in the museum sector, but also in other fields of application. The project thus impressively demonstrates the creative potential of artificial intelligence: it not only opens up new technical possibilities, but also creates innovative experiences – in the game, in the museum and at the interfaces in between.
