Alon Bochman’s Post

More news for people who allocate capital: An LLM with retrieval-augmented-generation (RAG) and fine-tuning for reasoning forecasts almost as accurately as a *crowd* of humans (71% vs 77%) Furthermore, if the LLM is allowed to only make forecasts when it is confident (ie when there is a lot of evidence, when the forecast is extreme, etc) the LLM is significantly more accurate than a crowd of humans. This is similar to Warren Buffet’s advice to “wait for a fat pitch” And, of course, when you average in the LLM’s forecast with crowd’s, you get even more accuracy. Again, the implication for capital allocators is clear: Institutional asset allocation is about to be disrupted. Paper: https://lnkd.in/d7C5Dy74 Previous post: LLMs make humans better at forecasting -- https://lnkd.in/ddVY6ZUi