You can find the investor interview with Marc Alexander Kühn from UVC Partners and all other episodes of our podcast on Spotify, iTunes, Amazon Music, Deezer, Google Podcasts, Pocket Casts, Radio Public, Breaker, Overcast, Castbox, Podcast Addict and Anchor.
Marc Alexander Kühn holds a Master's degree in Engineering and Computer Science from the Technical University of Munich with a research stay at the University of Oxford and a Bachelor's degree in Engineering and Management. He has also published several peer-reviewed scientific papers in the field of machine learning. His previous experience includes deep tech venture capital, computer science research at the Fraunhofer-Gesellschaft, technology strategy at a global company and freelance management consulting.
We start the podcast by briefly introducing UVC Partners. The venture capital firm behind the innovation center UnternehmerTUM focuses on disruptive B2B startups in the fields of deeptech, climatetech, mobility and software/AI. The geographic focus is on the DACH region as well as European teams looking to enter the German-speaking market. Now that the backer has Close fourth fund with 250 million euros the assets under management of UVC Partners now amount to over 600 million euros. Marc also explains what sets UVC apart from other investors. The dealmakers' technical background - he himself did research into machine learning - and UVC Partners' network - from universities to research institutions and industry - help to evaluate deep tech innovations.
The billion-euro gap in AI start-ups
From minute 5:25, we then turn our attention to the topic of AI. We start by talking about developments to date. Marc refers to the Gartner hype cycle and explains that start-ups now have to meet very high expectations and generate capital income. The focus is also on the reflections of David CahnPartner at Sequoia Capital. Ultimately, the issue here is that the AI start-ups would have to generate enormous profits with their software in order to finance the GPUs they need. There is still a huge billion-euro gap here and not every startup will be able to do this. That's why we then talk about how a VC finds the companies that can do this.
Then (from minute 10:20) we look at some of the technical aspects of AI. Marc explains in broad strokes how solutions like ChatGPT work. He goes into more detail about the Attention-model. This allows us to give context to the AI, from previous sentences to sources such as Wikipedia. We also discuss various problems that AI development may face in the future, from the origin of training data to increasing energy consumption (in the context of global warming). And Max also explains the difference or relationship between Machine Learning (ML) and Large Language Models (LLM).
Generative AI is not everything
We discuss AI use cases from minute 21:55 onwards, and it quickly becomes clear that technology and media-related industries are among the early adopters, while traditional industries with longer sales cycles will take longer to adapt. Marc points out that we should not forget that generative AI, i.e. solutions that generate texts, images or code, is only one form of artificial intelligence. Other forms, such as analytical approaches, also serve other use cases, for example in route planning or machine control.
Towards the end of the podcast (from minute 26:15), we also talk about the AI start-up landscape in Munich, Germany and Europe. Starting from the current German AI Startup Landscape we take a look at where Europe stands in the AI competition with the USA and what role Germany plays here. And we also take a brief excursion into the fields of robotics and automation - Marc's other areas of expertise - from minute 31:30 onwards. Here too, of course, AI plays a major role.
