I absolutely agree that history is filled with politically motivated bias. But there are actual historical facts (someone won the Siege of Vienna of 1529, and it wasn’t the Ottoman Empire). There are historical theories that actually fit most of the facts and pseudo-historical theories that fit carefully selected sets of facts. Being able to tell the difference is a valuable skill that members of this community should try to develop.
To put it differently, the falsity of the theory of moral progress has implications for assessing the difficulty of building a Friendly AI, doesn’t it?
There are historical theories that actually fit most of the facts and pseudo-historical theories that fit carefully selected sets of facts. Being able to tell the difference is a valuable skill that members of this community should try to develop.
And how does one do that? The problem is that most historical facts are publicly available, so how does one distinguish a theory producing by data mining and overfitting from one that wasn’t? The only historian I can think of who has anything close to an answer to that is Turchin via the usual statistics method of holding back data to test the extrapolations.
Turchin and Carrier are discussed occasionally, but not that much; why should I think this is not the right amount of discussion?
The bigger problem with most historical analysis takes the following form:
1) Pick a historical thesis (usually because it supports one’s pre-existing moral positions) 2) Find all historical evidence that supports that theory 3) Throw any remaining historical evidence in the trash
If you have successfully avoided that trap, congratulations. Society as a whole has not, and this community is not noticeably better than the greater societies we are draw from.
There are historical theories that actually fit most of the facts and pseudo-historical theories that fit carefully selected sets of facts. Being able to tell the difference is a valuable skill that members of this community should try to develop.
And how does one do that? The problem is that most historical facts are publicly available, so how does one distinguish a theory producing by data mining and overfitting from one that wasn’t?
I absolutely agree that history is filled with politically motivated bias. But there are actual historical facts (someone won the Siege of Vienna of 1529, and it wasn’t the Ottoman Empire). There are historical theories that actually fit most of the facts and pseudo-historical theories that fit carefully selected sets of facts. Being able to tell the difference is a valuable skill that members of this community should try to develop.
To put it differently, the falsity of the theory of moral progress has implications for assessing the difficulty of building a Friendly AI, doesn’t it?
And how does one do that? The problem is that most historical facts are publicly available, so how does one distinguish a theory producing by data mining and overfitting from one that wasn’t? The only historian I can think of who has anything close to an answer to that is Turchin via the usual statistics method of holding back data to test the extrapolations.
Turchin and Carrier are discussed occasionally, but not that much; why should I think this is not the right amount of discussion?
The bigger problem with most historical analysis takes the following form:
If you have successfully avoided that trap, congratulations. Society as a whole has not, and this community is not noticeably better than the greater societies we are draw from.
This is a thick problem.