When you get much smarter, you don’t realize that benefit by screwing extra pennies out of existing bets. You find new bets that you can now be 99% sure of instead of 50%, new things you can do with matter instead of decreasing tweaks to old things.
Yeah, I’m aware of that. I was Illustrating the returns to predictive accuracy in a single domain.
Cross-domain predictive accuracy wasn’t investigated here, and it’s not clear how much cross-domain predictive accuracy manifests in practice. That is, it’s not necessarily the case that if you improved your predictive accuracy in one domain (your predictive accuracy will also rise in other domains).
Are there any theoretical or empirical results suggesting that predictive accuracy generalises?
When you get much smarter, you don’t realize that benefit by screwing extra pennies out of existing bets. You find new bets that you can now be 99% sure of instead of 50%, new things you can do with matter instead of decreasing tweaks to old things.
Yeah, I’m aware of that. I was Illustrating the returns to predictive accuracy in a single domain.
Cross-domain predictive accuracy wasn’t investigated here, and it’s not clear how much cross-domain predictive accuracy manifests in practice. That is, it’s not necessarily the case that if you improved your predictive accuracy in one domain (your predictive accuracy will also rise in other domains).
Are there any theoretical or empirical results suggesting that predictive accuracy generalises?
Google’s GATO perhaps? I’m not sure in which direction it actually points the evidence, but it does suggest the answer is nuanced