Björn Ommer nominated for Deutscher Zukunftspreis

Enlarge

Through the Deutscher Zukunftspreis (German Future Prize), the President of Germany recognizes outstanding technical, engineering, or scientific achievements or software- and algorithm-based innovations which significantly expand the international state of research and technology and are already in practical use. The prestigious award comes with a prize of 250,000 euros.

Today, on 11 September 2024, three projects were presented for the final round of the Deutscher Zukunftspreis at the Deutsches Museum in Munich. One of them is the project “Democratizing Generative AI – Stable Diffusion from Development to Practice,” which Professor Björn Ommer presented as spokesperson. “My team and I feel very honored by this nomination,” says the AI expert and Chair of AI for Computer Vision and Digital Humanities / the Arts at LMU. Before being elevated to this distinction, all three candidates went through a process of rigorous examination. As the first step, prominent German science and trade associations put forward innovative developments for the award following intensive analysis. And then a ten-member jury assessed the scientific achievement and economic significance of the proposed projects. Stable Diffusion passed this test and is now in the top 3 in the race for the Deutscher Zukunftspreis crown in 2024.

Democratizing generative artificial intelligence

“Generative AI is about getting a computer to learn how to generate content – such as images,” explains Ommer. The goal is to capture local details of an image and its global context. For AI to be able to learn these relationships from training data, it needs a very large artificial neural network. But therein lies the catch: Such networks require large, expensive computing capacities in practice. As a consequence, only large tech corporations have the resources needed to further develop such AI models. “With Stable Diffusion, we want to dissolve the resulting dependencies and create AI models that are just as effective but require much less computing power in their application,” says Ommer. And so his chair has undertaken the task of democratizing generative AI. Through its application by the company nyris GmbH under the leadership of Dr. Anna Lukasson-Herzig, this undertaking is fundamentally transforming visual search and thus substantially increasing productivity.

To minimize the memory and processing costs, the researchers devised an innovative approach whereby the AI first learns a new, effective image description language for local image regions instead of directly describing images in terms of individual pixels. “As this compact and efficient generative AI can become a catalyst for countless applications, we think it’s important to make the software free with open access to all,” says Ommer. The team is working on further developing Stable Diffusion and creating new applications. “In the long term, the goal is to expand the possibilities of generative AI and above all to render the communication between humans and machines more efficient,” adds Ommer. “There’s tremendous potential here for all our futures and it’s incumbent on us to exploit it.”

The presentation of the Deutscher Zukunftspreis and the announcement of the winners will take place in Berlin on 27 November 2024.