Summary

Laurent from AMS, a data provider specialising in large language models (LLMs), discussed the challenges of inflation forecasting, particularly in the short term, due to sector-specific idiosyncratic shocks and limited high-frequency data. He emphasised the difficulty of obtaining quality input data, especially for services, and described how AMS uses LLMs to track and aggregate inflation-related news. Their system scans hundreds of news sources globally, translating and summarising data to build indices and time series that track inflation trends in various sectors like energy and food. Laurent highlighted the use of diffusion indices, which measure the balance between positive and negative news to indicate inflation trends. He presented insights from their models on U.S. core inflation, noting stable overall momentum but with sectoral disparities. While insurance-related news showed significant upward pressure, other core components like airfares and car prices exhibited weaker trends. He concluded by expressing caution about potential near-term inflation volatility, despite some stabilisation in core inflation data.

Topics