Have any experiments been done with AI to put cyclical theories of history to the test? What if any were the results ?
The European Science Foundation made and has made a big push for quantitative methods in the social sciences, obviously including machine learning and other methods that the researchers see fit to apply, but it is a big field and unfortunately I cannot tell you much off the top of my head.
You may or may not be aware of the recently released books by Peter Turchin, who calls what he does cliodynamics [clio- being the Greek muse of history], with big data sets being examined statistically and mathematically. His topic is the rise and fall of complex societies.
His most recent book has been released recently and if you type his name you’ll find many gushing reviews in the places where people gush about things. You’ll find much more considered reviews by people who are experts in history in the mainstream press. Its fresh enough that some of these are still paywalled.
I’ve not read any of his work, but people whose opinions I would tend to trust (academic historians and people who do this for a living and think about it as their job) are not as excited as the people who find normal history dull and think this will [a] sexify it up a bit and [b] let you monetise history and the future.
I won’t offer an opinion on something I’ve not read, but generally human history is big enough that you can find anything you want to start looking for, especially if you have to code data in certain ways to begin counting it. I don’t have any sense as to whether Turchin’s data is as value-free as it might need to be. The bits cited by reviewers incline me to think his analyses are glib and skip over necessary complexity.
You might also to search up <Karl Marx - historical cycles - big data> or combinations like that. There were a few interesting looking references.
Hari Seldon is working on it. But I don’t think he has published anything yet?
I still don’t know what possible meaning the sentence “Have any experiments been done with AI to put cyclical theories of history to the test?” has.
AI can’t do experiments with history. It could churn out 10,000 responses to any question involving “cyclical theories of history” but none of them would be any more correct than the others. And who, as noted above, would be qualified to definitely determine the “correctness” of any answer?
Quite true. We need the ‘alternate-history-o-tron’ machine for that.
Unfortunately the temporal convolution capaciter in mine is broken, and you can’t get the parts here and now: the unobtanium doesn’t exist in our periodic table.
Note to moderators: this really shouldn’t be in the ‘factual questions’ category, I think?
Summoning @Hari_Seldon…
To even begin to accurately simulate the thoughts and actions of large numbers of human beings would take multiple orders of magnitude more computer power than currently available, complete understanding of how human brains work down to the atomic level, and a near omniscient level of knowledge. So I’m guessing “no”.
I work only on acyclic models.
A circle has no end.
But it also doesn’t have a beginning!
I think that the idea is that you take all of the historical data we have, pour all of it in a machine, and let the machine try to tease out all of the patterns in it.
The problem is that we probably don’t have the relevant data, at all, for most of history. Historians of the past have written plenty about how, for instance, daily life changed as a result of the fall of the Roman Empire, but there was much less recorded about how things were changing just before the fall of the Empire, because nobody at the time saw any need to record all of that. And that’s the sort of data we’d need to be able to predict the fall of a modern empire.
Or you could just use “all of that portion of history for which a significant newspaper record survives” as your input set. There, we probably do have the data. But now your data doesn’t go back far enough to encompass a lot of sorts of important events, like the fall of a major empire. So you still can’t predict things like that.
That is not how it works. You have to work with indicators and data you can at least in principle try to quantify, like economic status, demographic structure, social support, interest rates, anything quantifiable.
That’s what I was wondering.