The pipeline is very simple: Substrack RSS -> "analyze for insights" prompt -> insights table in db. Hallucinations limited by "quote the text verbatim" prompt. If I'm not mistaken I am using Gemini for this but I actually forgot.
We are doing an event at Climate Week NYC this year, and looking for ideas for speakers that have insights to share from their climate work/research/activism efforts. We are interested in technical perspectives, activist perspectives, policy perspectives, all of it. The positive, the critical. We have a preference for diverse speakers. Who should we reach out to?
1. Steven Koonin - professor at NYU, Under Secretary for Science at the U.S. Department of Energy in the Obama administration.
2. Michael Shellenberger - an author and journalist who writes about politics, the environment, climate change, and nuclear power. He is a co-founder of the Breakthrough Institute and the California Peace Coalition.
3. John Clauser - Nobel prize in Physics (2022) and Wolf prize in physics (2010) recipient.
4. Judith Curry - President of Climate Forecast Applications Network (CFAN) and former professor and Chair of the School of Earth and Atmospheric Sciences at the Georgia Tech.
An additional (and I was quite surprised by this) trick is to ask an LLM to reformulate the user prompt “to be more concise and precise”, then run a vector similarity search against that, which (in our experience) leads to better matches.
There are many tricks to get better context to send to your LLM, and that’s a large part of making the system give good answers.
I built an app to give us analytics on an LLM we are building for a client. It focuses specifically on “what people are talking about”. It’s literally one call to Gemini in terms of LLM usage… https://simplyanalyze.ai/ (We worked on this after the initial weekend but the basic app and UI you see there was working in a weekend)
