Earlier quoted context omitted.
I wouldn't put a lot of faith in such predictions either, but there are a few reasons to believe that exceptionally good returns might not be repeatable. One is simply regression to the mean. Others are high global levels of debt and a significant wave of globalization having just run its course. But I think productivity growth and our ability to redistribute the fruits of that productivity growth are the big unknown…
> The current step change in our AI capabilities could throw a spanner in all attempts at making any reliable predictions. On the other hand, vastly improved AI capabilities could mean that future prediction systems are more reliable. At some point in the future, all further predictions may become roughly accurate. Maybe these AI sytems are clever enough in make only predictions which are self-fulfilling prophecies.…
Our current AI systems are pretty good at identifying patterns of past behavior. But the extent to which past behavior contains information about future behavior may be limited. Information that isn't there cannot be found irrespective of how clever any predictor may be.