
KEYNOTE - Ronald Ashri
OpenDialog.ai
Designing Software that Acts: Agents beyond LLMs
About the talk
While LLMs have renewed interest in autonomous software agents, current implementations often overlook decades of foundational research. The end result is that we keep re-discovering the same problems while we could be leaping ahead. The aim of this talk to to bridge this gap by examining how established agent theories and architectures can enhance today's LLM-driven applications.
We dive into the history of multi-agent systems research and explore how frameworks such as Belief-Desire-Intention or different coordination mechanisms can provide valuable insights of developing today's systems. By combining historical agent paradigms with modern AI capabilities, we can move beyond prompt-centric development toward robustly engineered agent systems.
By the end of the talk you will have a clearer understanding of what an agent-based system is irrespective of whether you are using LLMs and a more nuanced perspective into how LLMs can be used in the context of a multi-agent system.
About the speaker
Ronald is a technologist and entrepreneur with an academic background and over 20 years of experience in product development, artificial intelligence, and building and growing technology companies.
His current work focuses on applying AI and emerging technologies to solve real-world business problems as a Co-Founder and the Chief Product and Technology Officer of OpenDialog.ai
He has authored several books and academic papers on AI and multi-agent systems and write/speak frequently on the topic.
He runs a Substack newsletter called AgentsDecoded.com where he shares his thoughts on AI agents and multi-agent systems.