If you use frequentist statistics you can just claim that your data follows a normal distribution (which it objectively most likely isn’t, even the archtypical example of height violates a normal distribution because there are more people with dwarfism than the normal distribution assumes). On the other hand, to use Bayesian statistics you do need to decide on priors which does involve subjective decisions.
Objective in a sense, but I’m not sure how I’ve given you the ‘God’s view’ impression. I think epistemology should be objective in that it should work universally via the same rules, and that people can discuss both ideas and the world (evidence) in such a way to reach agreement in an objective sense.
The fact that you aren’t explicitly thinking of God doesn’t mean that the idea of objectivity does not come out of a Christian scientific tradition which had as a key motivation trying to see things from God’s view. Your ideas of how you think, it should work have that theistic origin.
Bayesianism as discussed on LessWrong is about how it would be good for an agent to reason and that’s a different goal. Eliezer was interested in it because he wants to know how what’s true about how agents effectively reason to be able to say things about superintelligence.
Philip E. Tetlock separately was interested in the epistemic task about how to reason about what to believe. After the Iraq war the US military thought they had problems with epistomology as shown by the fact that they got the WMD question so horribly wrong. Out of that there was an IARPA tournament where Tetlock’s team won. Tetlock runs GJOpen which does solve epistemic problems for entities that want probability for certain events happening. He got some grant money from OpenPhil. There’s also Metaculus that comes out of rationalist sphere. His book Superforcastingis a good resource if you want to understand how the kind of Bayesianism where people don’t explicitly use statistics works in practice.
If you use frequentist statistics you can just claim that your data follows a normal distribution (which it objectively most likely isn’t, even the archtypical example of height violates a normal distribution because there are more people with dwarfism than the normal distribution assumes). On the other hand, to use Bayesian statistics you do need to decide on priors which does involve subjective decisions.
The fact that you aren’t explicitly thinking of God doesn’t mean that the idea of objectivity does not come out of a Christian scientific tradition which had as a key motivation trying to see things from God’s view. Your ideas of how you think, it should work have that theistic origin.
Bayesianism as discussed on LessWrong is about how it would be good for an agent to reason and that’s a different goal. Eliezer was interested in it because he wants to know how what’s true about how agents effectively reason to be able to say things about superintelligence.
Philip E. Tetlock separately was interested in the epistemic task about how to reason about what to believe. After the Iraq war the US military thought they had problems with epistomology as shown by the fact that they got the WMD question so horribly wrong. Out of that there was an IARPA tournament where Tetlock’s team won. Tetlock runs GJOpen which does solve epistemic problems for entities that want probability for certain events happening. He got some grant money from OpenPhil. There’s also Metaculus that comes out of rationalist sphere. His book Superforcasting is a good resource if you want to understand how the kind of Bayesianism where people don’t explicitly use statistics works in practice.
I’m not sure what the best shorter source is but maybe https://www.gjopen.com/training/