Forecasting world events with language models, with Shashwat Goel

​Forecasting world events with language models
Talk by Shashwat Goel, Max Planck Society
Hosted by Daniel D'souza, Research Engineer, Cohere Labs

​Can language models predict the future? I'll discuss our ongoing work on the OpenForecaster project. I'll start by discussing challenges in evaluating language model forecasting abilities. Then I will describe a scalable recipe that uses daily news to post-train language models for producing calibrated forecasts of world events. Finally, I will present our recent results on how we can replay world events as they happened to measure frontier agents' ability to learn from new information and continuously adapt their forecasts.

This talk is part of Cohere Labs in Conversation, a limited series of talks, in which Cohere Labs scientists and engineers host external researchers for techincal talks and Q&A discussions on subjects related to our current explorations at Cohere Labs. We look forward to sharing these talks with you, giving you a glimpse into the problems we're exploring, and learning together from some of the greatest minds in the field.