Daniel Erasmus has spent about 25 years doing scenario planning for banks, governments and companies, and his starting point is that experts are bad at predicting the future. In 2006, a scenario he worked on helped one bank prepare for a global financial collapse. When the crisis hit in 2008, it was one of only two banks in the world to grow its assets.
We recorded this conversation at DWeb Camp 2026 outside Berlin. Daniel explains how scenario planning works and how scenario work during COVID flagged a food security risk, which led a bank to start a program that has since given away about 120 million meals. We also talk about ClimateGPT, the open climate model he built, why he thinks climate is a bigger danger than AI, and how he answers the critique that AI uses too much energy.
Daniel Erasmus is head of AI at Internet Archive Europe and a member of the Club of Rome. He built ClimateGPT, an open source and open access AI model focused on climate and the social decisions around it.
0:00 Intro1:42 What is Scenario Planning?11:25 The Scenario Planning Process & Technology27:04 What is Climate GPT?34:37 Why Use AI for Climate? The Degrowth Question45:06 The Role of Internet Archive50:03 Climate Economics & Negative Externalities
This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit allmodelsarewrong.online/subscribe
Podden och tillhörande omslagsbild på den här sidan tillhör
Joshua Davila. Innehållet i podden är skapat av Joshua Davila och inte av,
eller tillsammans med, Poddtoppen.
