That’s it Sam. That’s what makes it awesome. If you have some process that you know nothing about, this is your formula. No surprise that it produces wide bounds.
Simple answer: it’s a great starting point for your analysis. It’s a baseline. Now you can design a more detailed research project.
Which is what happened.
Yes, once we get started. Gott’s first paper took humanity’s current duration and used it to extrapolate their future duration. The next step is to note that the odds of extinction jumped up during the 1940s, especially after Hiroshima. So apply the same formula to years after 1945. The step after that is to think methodically about possible extinction risks: that’s what eg Nick Bostrom’s group devoted themselves to. Along the way he/they created the Grand Simulation thought experiment, which we’ve covered in GD before (though I couldn’t find the discussion I was looking for - I see erislover hasn’t been around since 2016).
I’d also argue that we should think more deeply about the data generating process, how one year’s risk is correlated with the next year’s risk. Even if you can’t adequately measure a risk’s level, you might be able to get some traction from looking at it’s sequencing. Or not.
Think of it this way. If you truly believe that you live in a dynamic system where small pertubations in the initial conditions change everything, then basically all you have to work with is the 1/3 - 3 times rule, or a more sophisticated version of it. But as you said, usually there are other processes involved, many of them indicating some sort of equilibrium, that allows you to put more structure on the problem and reduce the error bounds.
Back to the OP. The industrial revolution gets a lot of PR, but economic growth rates didn’t ramp up higher until around 1870, when the communities of competence surrounding the modern corporation emerged. R&D. Marketing. Engineering. We’ve had just over 150 years of this new era - a short dataset.
A bit about data
US economic datasets emerged during the Great Depression and shortly after WWII. Some US and UK data goes back further. UN and OECD data generally dates to 1960. All this is helpful, but it ain’t psychohistory.
Imagine if you had instead a million years of data. And also thousands of star systems, each with hundreds of governments historically speaking. Allowing for some artistic license (a lot of artistic license) you could have something like psychohistory. As it is, we are stuck with cliometrics.
Still think predictive models are impossible? Decent weather forecasts out to 10 days are now available for free and have been for years. 30 days, not so much.