I remember reading about a company that had these 5-year projections of what their sales and revenue was going to be. A new CEO came in and said, wait a minute, we can't accurately predict what our sales are going to be next month. How are we going to make 5-year predictions that are anything other than worthless?

He scrapped the long-term planning system, shrinking the planning window down shorter and shorter. They ran the company on a 30-day window for a year or 18 months. Only then did they get the numbers right. After that, they extended the planning window, but never back to five years.

So, those who are doing longtermism: How accurate are your short-term forecasts? Can you accurately predict a year from now? If not, how can you predict the long term future with enough accuracy to act on?

Worse: For every year further in the future, you should probably multiply your certainty in your prediction by 0.8. (This depends on the nature of the prediction, of course, and is a made-up number. Still, the point is valid - the longer term the prediction, the more probability that it is not only wrong, but wildly wrong.) That means (using that number) that predictions 10 years out only have 10% accuracy. Predictions 20 years out only have 1% accuracy. How do you think you know enough today about what will happen then in order to make decisions now on the basis of what will happen then?