#rp23 keynote speaker Björn Ommer: What it means to teach computers to "see".

19.04.2023 - Björn has developed the generative AI model "Stable Diffusion". Despite its great transformative power, he calls for critical dialogue.
Björn Ommer im Anzug vor einer weißen Wandtafel
Photo Credit
Fabian Helmich / LMU

Together with his research group, Björn Ommer is teaching computers to "see" – i.e. to understand and learn from our world simply by watching it. Although true image understanding is still an ambitious goal, this research has already demonstrated its great transformative power: recently, the group has published a generative AI known as "Stable Diffusion", which is now democratising the creation of visual content and opening up new directions in the arts, media, entertainment, and beyond. 

He is fascinated to see what we will be able to create with improved support by computers in the future, when AI is turning them into more powerful and more accessible tools. Björn also sees a great need for public dialogue on how societies can positively embrace this powerful technology to maximise the benefit for the masses and minimise its risks.

Björn Ommer is a full professor at the University of Munich, where he is heading the Computer Vision and Learning Research Group. Prior to this, he was a full professor in the department of mathematics and computer science at Heidelberg University. Björn received his diploma in computer science from University of Bonn and his PhD from ETH Zurich. Thereafter, he was a postdoc at UC Berkeley.

At #rp23 we look forward to Björn's fascinating insights into the field of AI and computer vision.


An interview with Björn Ommer.

Generative AI appears as a great revolution right now. But is it really?

Yes, unlike previous hype around AI, we are now seeing it transforming many applications and thus enabling new technologies. Functioning prototypes are already being used by many millions of people. Even if these still have a lot of potential for improvement, they already represent a huge bonus for many.

What questions should we ask ourselves as a society when we talk about (generative) AI?

It is a very powerful technology with great potential – as an enabler for other technologies, but also with the possibility of negative uses. Its development is taking place worldwide and will therefore hardly stop. We therefore need a broad public discussion on how we want to use generative AI in the future – and how not.

Stable Diffusion, which you co-developed, is open source. Why did you and your team take this step?

The development of generative AI is internationally particularly being driven by large companies. Previously, the models were computationally so intensive, that their application and further development seemed increasingly limited to a few large companies with large data centres. At the same time, when we published "Stable Diffusion" last year, the technology was not yet qualitatively advanced enough for the images generated to be commonly considered as “real”. Therefore, it seemed to be the appropriate time to democratise research on this technology and – at the same time – raise awareness of its uses and potential consequences among the general public.