Motif Clarity – A Temporal Causal Knowledge Graph to Better Connect the Dots

Session track: Data Intelligence

Session time: 04:00 am - 04:30 am UTC November 12

Session description:

We started motif - an ai assisted wealth advisory platform - with the goal to enable people globally to invest with confidence. Originally we tried to use a thin wrapper around LLMs and some workflows around it to understand what's happening in the world and how that might impact an investment, but we quickly realized that that was not working. LLMs are not grounded, they hallucinate, they do not understand causality and temporal relationships per se. Recently there has been a lot of buzz around knowledge graphs, but we thought most of them still lack two concepts: causality and temporality. We wanted to understand: what events lead to the situation we are in now, what are the events that might happen down the line and whats the relationship between these events and a company. A practical example of this is the events that lead up to the war in iran, the supply chain implications and what this means for various assets like crude oil, technology stocks and so on. In this session, we will hands on explore the clarity pipeline to create the causal, temporal knowledge graph and then the agents that extract knowledge from this. We will look at taxonomy, strategies to create a manageable, sparse knowledge graph for the agents, pitfalls and wins.