The second TACHELES Conference, organized by the FZI Research Center for Information Technology on June 18, 2024, was all about Generative AI for autonomous and connected driving. Around 100 participants gathered at the CyberForum in Karlsruhe to listen to and engage with exciting presentations, participate in panel discussions, expand their networks, and work on specific issues in workshops focused on key topics of generative AI for AVF. It was an intensive conference day for everyone involved.

While the first edition of TACHELES – the alternative conference on autonomous and connected driving – last year focused on the homologation of Level 3 and Level 4 systems and vehicles, this year's event addressed the omnipresent topic of generative AI and its potential impact on Level 4 and Level 5 autonomous and connected driving. The promises associated with generative AI are huge, raising the question of whether it could revolutionize autonomous driving. The conference explored whether we might be facing yet another hype or if this universal technology offers crucial new opportunities necessary for the development of higher-level automated driving.

Presentations with Fascinating Insights
Renowned tech blogger Sascha Pallenberg delivered a keynote showcasing the impressive technical advancements in computing power over the past 15 years, highlighting what current AI and chip technologies can achieve – from fully automated production processes to autonomous driving and global weather forecasting models. The subsequent technical presentations by Axel Gern and Jan Zawadzki focused on autonomous driving. They emphasized how generative AI could influence further development and support the European automotive industry's promise to bring safe autonomous driving to the roads, noting that players in China and the USA are currently reaching technical limits.

Controversial Panel Discussion with Audience Participation
These topics and more were the focus of the following panel discussion, which featured additional experts Tille Rupp from Porsche Engineering, Rebekka Görge from Fraunhofer IAIS, and Prof. Tim Fingscheidt from TU Braunschweig. Philipp Franke, representing the Baden-Württemberg Ministry of Transport and patron of TACHELES, introduced the political perspective to the discussion. The audience also had the opportunity to contribute numerous questions and additional insights during the nearly one hour program.

Intense Workshops on Metrics, Simulation, and Frameworks
In the afternoon, participants engaged in three themed workshops. The TACHELES program advisory board set the agenda with topics on generative AI for autonomous and connected driving: Metrics ("Who measures, measures wrong"), Simulation ("Generative AI for Simulations in Autonomous Driving: Requirements, Potentials, and Limits"), and Frameworks ("Successfully implementing AI in the automotive environment from a technical and regulatory perspective"). The Simulation workshop received particular attention due to the promising possibilities of generating synthetic data with generative AI, allowing for the creation and testing of traffic scenarios at will, which is a hot topic in both industry and academia. This workshop also highlighted the significant need for common standards for the broader use of simulated data.

In the subsequent workshop panel, participants were able to gain an overview of the approaches taken in the workshops, identified pain points, and aha moments. Initial ideas for recommendations for action were also presented, with a call to build on these important discussions and continue them – whether in research projects, working groups, or interdisciplinary challenges.

Visibly satisfied, Marius Zöllner, the FZI host, concluded the second TACHELES – the alternative conference on autonomous and connected driving. He advocated for a next edition in 2025 and encouraged the submission of topic suggestions.

Once again, EICT organized the conference for the FZI, from conception to planning and program design to on-site organization.

More about the event