How does artificial intelligence actually learn? This question is addressed by an interactive demonstrator from inIT in the special exhibition ‘A KInd of Art. Artificial Intelligence Meets (Weser) Renaissance’ at the Weser Renaissance Museum in Brake Castle. The demonstrator clearly shows how AI models recognise patterns from data and use them to make predictions. It was developed by Nico Baumgart, a research associate in the Mathematics and Data Science working group headed by Prof. Dr. Markus Lange-Hegermann.
How artificial intelligence learns and where its limits lie
The demonstrator, entitled ‘Your age. Your height. Model meets reality,’ uses a low-threshold example to show how AI models learn from data and make predictions based on it. Visitors can add data points themselves – such as age and height – and observe how this changes the models' predictions.
Two different Gaussian processes are trained and visualised. They show how an AI model recognises correlations and estimates future values based on existing data. The more – or better matching – data is added, the more accurate the predictions become.
Learning with uncertainty
It becomes particularly interesting where little or no data is available: in these areas, the uncertainty of the models increases significantly, which shows how much artificial intelligence depends on the quality of its data basis.
‘Artificial intelligence can do a lot, but it is not the solution to everything,’ emphasises Nico Baumgart. ‘Ultimately, it is up to us to train models sensibly, use them responsibly and interpret their results correctly. Because only with high-quality data can reliable predictions be made.’
Special exhibition until mid-December
With the special exhibition ‘A KInd of Art’, the Weser Renaissance Museum at Brake Castle builds a bridge between the past and the future. It can be seen at Brake Castle in Lemgo until 14 December 2025. Visitors can discover various facets and applications of artificial intelligence.
Further information is available at:museum-schloss-brake.de/portfolio_page/akindofart/