Forecasting the Impact of Artificial General Intelligence (AGI)
In 2017-18 Foresight published a five-part series on “Forecasting the Impact of Artificial Intelligence” by Spyros Makridakis. In a follow-up assessing his earlier forecasts – and looking ahead to the rapid advancement of artificial general intelligence (AGI) – Spyros warns us that AGI can become our greatest partner, or our greatest peril.
Special Feature: TimeCopilot
Agentic Forecasting with TimeCopilot: Unifying LLMs and Time Series Foundation Models
Agentic forecasting mirrors how human experts work, using large language models to plan, select, evaluate, and explain forecasting workflows. Azul Garza and Renée Rosillo show how TimeCopilot integrates various methods into a single interface, eliminating the need to manage multiple modeling pipelines.
Commentaries
- Christoph Bergmeir examines the use of LLMs in forecasting
- Anne-Flore Elard discusses the implications for teaching forecasting
- Zabiulla Mohammed explores making advanced forecasting more accessible
- Fotios Petropoulos applauds the great opportunity in agentic forecasting while identifying some challenges.
The Cost of Forecast Error Revisited with Conformal Prediction
Peter Maurice Catt revisits his 2007 paper “Assessing the Cost of Forecast Error.” In the new piece, Peter considers critiques of his earlier framework that relied on classical normal distribution assumptions. His updated approach utilizes conformal prediction, creating a modern framework using distribution-free methods.
Retraining as Approximate Bayesian Inference
Model retraining is usually treated as an ongoing maintenance task. But Harrison Katz argues that retraining can be better understood as approximate Bayesian inference. He describes the gap between a continuously updated belief state and a deployed model as “learning debt,” and the retraining decision becomes a cost-minimization problem.
Categorical Forecasting for Garage Management
Aris Syntetos and Shixuan Wang investigate a new application in Foresight: forecasting automotive repair times for garage management. They find that cost-effective job scheduling relies upon assigning repair times into appropriate categories, rather than using exact repair time lengths.
Essay ~ The Quiet Skills behind Forecasting
Jerry Raphael, who comes from a background in global agricultural markets, argues that demand forecasting is more than a statistical task and emphasizes the quiet skills behind forecasting – making it a discipline for understanding how complex markets move.
Access Foresight
- IIF members click here to access the current issue and archives.
- Non-IIF members click here to access the current issue only. To become a member, click here.
