We're in a period where the decisions society writ large, and a few individuals in particular make, will be very consequential for the future of humanity. So, it matters that we get those decisions right.
If we can have the technology to make better forecasts and plans, and to navigate a chaotic information environment, we'll likely all make better decisions, and more good things will in fact happen.
Principled forecasting has long been a quixotic quest of mine. People making rigorous plans, and forecasting the outcomes of those plans feels like it should be the thing by which we steer the future, and in fact, there is strong evidence that superforecasters, forecasting aggregators like metaculus and manifold markets, and real money prediction markets, can produce calibrated forecasts on important events.
However, despite many attempts, bringing quantitative forecasts into the planning process for projects has not been very successful. They do not seem 1.) flexible enough or 2.) comprehensive enough or 3.) something! to meaningfully shape the planning process, and thus the decision making process.
I had given up on 'principled forecasting' as an intellectually attractive but ultimately distracting enterprise; however, like in so many other domains, advances in AI might have changed the calculus. There's potential to augment so much of the forecasters workflows:
I also suspect we're already able to forecast somewhat effectively directly with AI and that this will improve fast. If researchers and technologists can build the forecasting systems that are actually useful it would be incredibly impactful