ETA: I misunderstood Raemon’s comment—see his reply.
Of course it’s helpful information. I’m not claiming that everything about EA Funds is bad, and I’m pretty annoyed at this pattern where I’ll make a particular criticism and people will respond to some other criticism. I’m specifically claiming that there isn’t really info about track records with respect to outcomes, despite this being a large portion of the basis on which EA is marketed.
I was just responding to the “this will change nobody’s mind whatsoever” bit. I was someone who had some vague sense of “not sure what the deal with the funds is, leaning at this point towards “they’re probably not a good place to give money”, and having someone do some legwork of checking up on that was helpful (for the purposes of changing my mind)
In fairness, it’s only been a year and some of these may take longer to have reasonable track records. But if so, there should ideally be reporting on proximate targets, and clear indications of what the endpoint is and how it might eventually be measured (or if it can’t be, a clear accounting for correlated prediction errors which will never be corrected).
Oh, oops—I read your comment backwards. Thanks for clarifying! Sorry I was a little oversensitive this time, gonna try to update on the fact that this was a false positive :)
ETA: I misunderstood Raemon’s comment—see his reply.
Of course it’s helpful information. I’m not claiming that everything about EA Funds is bad, and I’m pretty annoyed at this pattern where I’ll make a particular criticism and people will respond to some other criticism. I’m specifically claiming that there isn’t really info about track records with respect to outcomes, despite this being a large portion of the basis on which EA is marketed.
I was just responding to the “this will change nobody’s mind whatsoever” bit. I was someone who had some vague sense of “not sure what the deal with the funds is, leaning at this point towards “they’re probably not a good place to give money”, and having someone do some legwork of checking up on that was helpful (for the purposes of changing my mind)
In fairness, it’s only been a year and some of these may take longer to have reasonable track records. But if so, there should ideally be reporting on proximate targets, and clear indications of what the endpoint is and how it might eventually be measured (or if it can’t be, a clear accounting for correlated prediction errors which will never be corrected).
I am now laughing at myself, because independently I read your comment the way Ben did, and downvoted it. (Have now removed the downvote.)
Oh, oops—I read your comment backwards. Thanks for clarifying! Sorry I was a little oversensitive this time, gonna try to update on the fact that this was a false positive :)