Focus on multi-agent systems

More and more applications are not just relying on a single AI - but on networks of several autonomous agents: so-called multi-agent systems (MAS). These enable different AI agents to take on tasks simultaneously, communicate with each other and achieve complex goals together. Particularly interesting for industry, production and knowledge management: MAS offer the potential to make processes more efficient, flexible and scalable.
The AI Network meeting addressed precisely these developments. Participants gained insights into current research and practice and learned how such systems are technically implemented. There was also an opportunity to discuss issues relating to security, trustworthiness and future prospects.

Cybersecurity as the basis for trustworthy AI agents

Prof. Dr. Bernhard Bauer from the University of Augsburg began the network meeting with a presentation entitled "Trustworthy autonomy: cybersecurity as the basis for trustworthy AI agents". He explained how multi-agent systems can be designed in such a way that they not only function efficiently, but also securely and trustworthily. He looked at architecture, communication and control mechanisms - particularly with a view to preventing cyber attacks and uncontrolled interactions between agents.

Taking robotics to the next level with multi-modal LLMs

Fabian Ritz from MaibornWolff GmbH then presented "How to take interactions with a robot arm to the next level with multi-modal LLMs". He showed how modern, multimodal AI models can be used to control and interact with physical devices - specifically a robotic arm. In doing so, he demonstrated that MAS and AI are not just abstract software concepts, but also enable concrete industrial applications - for example in automation, robotics and manufacturing. His presentation made it clear how closely AI, sensor technology and robotics can be interlinked and how this can create new forms of interaction with the real world.

Using multi-agent systems safely

In the third part, Wolfgang Graßhof from AuxData.ai spoke about how multi-agent systems can be used safely. He discussed the requirements that need to be met when planning, setting up and controlling MAS to ensure that they run reliably and deliver high-quality results. This included questions such as: When and how should agents be started? How can their interaction and performance be managed? And how can errors or unwanted side effects be avoided? His presentation provided guidance for companies and developers in particular who wanted to use MAS in practice - with a focus on usable, secure and sustainable implementation.
Taken as a whole, the three presentations offered a balanced review of multi-agent systems: they ranged from fundamental security and trust issues to specific robotics applications and practical implementation in everyday corporate life. They made it clear that MAS are more than just a technical vision - implemented correctly, they can create real added value.
The network meeting was rounded off in the usual way with a networking session - a good opportunity to make contacts and reflect together on application scenarios from research and practice.

Are you interested in the topic of multi-agent systems (MAS)? Are you an SME and wondering whether you could use MAS in your company? Are you an expert and would like to share your knowledge at the next AI Network meeting? Would you like to discuss the topic of AI in production? Feel free to contact us!