AI Signals #23 Reasoning Models, Scaling Workloads, and Beating Bias

​Agenda

​06:30pm - Doors open, food and drink served

​07:00pm - Welcome

​07:05pm - Chaitanya Pathak, Leads by Analyttica "The Evolution of Reasoning from System 1 to System 2 Thinking in Large Language Models (LLMs)"

​The shift from System 1 to System 2 thinking is essential in overcoming some of the inherent limitations of current LLMs, such as susceptibility to bias, overgeneralization, and errors in complex tasks that require nuanced understanding or multi-step reasoning. In this session, we will explore how LLMs are transitioning to more sophisticated forms of reasoning, including how they handle ambiguous queries, adapt to novel contexts, and generate more thoughtful, context-aware responses.

​07:45pm - Break

​08:00pm - Ben Gamble, Ververica "Scaling AI workloads with Flink."

​Ben’s talk will run though why Flink is a great tool for handling AI powered services and especially Agentic systems. He will talk about data, events, and semantics.

​​08:40 - Damien Jade Duff, Daemon "Beyond multi-agent systems: well-designed AI systems delivering value"

​Multi-agent systems are currently trending and they do have their uses and an interesting future, but equally they are being thrown with little thought at problems for which modern AI approaches that have been carefully designed and well-tested. Here I give our experience of use cases where we have used good practice and wit to orchestrate LLMs for use cases around knowledge distillation, data extraction, and controllable customer-facing solutions.

​09:00pm - Wrap up, drinks at Angel London