Neural Notes - Are more LLM calls all you need?

On this episode of Neural Notes, Chase Roberts and Sandeep Bhadra of Vertex Ventures US interview Lingjiao Chen, PhD candidate at Stanford University, on his team's research into developing a scaling law for compound AI systems.
In short: More AI calls doesn't always equal improved performance, and being able to predict the ideal number of model runs for a query would help make compound AI systems both more performant and more accurate.
You can read the paper we discuss here: arxiv.org/pdf/2403.02419
#startups #ai #artificialintelligence