2023-09-08T05:28:49+00:00 | 🔗
@timschlomi 1) with enough effort, yes. But well-calibrated models are nowhere near useful. Simply saying the mean is "well-calibrated" but not useful 2) not possible, I know enough about trying to beat expert experts to know not possible
2023-09-08T05:27:18+00:00 | 🔗
@timschlomi But the claim would be that prediction markets are better than pleas? I think given the uncertainty around the accuracy of prediction markets that's not clear and I would say no Aside, Would you allow pleas in this prediction market system?
2023-09-08T05:23:55+00:00 | 🔗
@timschlomi people insist it's mathematically secure and therefor it's secure. But in practice, I mean just look at the evidence it's not very secure. I think prediction markets are the same way.
2023-09-08T05:23:13+00:00 | 🔗
@timschlomi Again I'm much more confident in how well-calibrated they are than how accurate they are. I think that this is the path towards needing to agree on a common single definition of "Justice" and opens up our justice system to too many attack surfaces. It's a bit like crypto...
2023-09-08T05:21:41+00:00 | 🔗
@timschlomi I think it gets tricky to compare ideal systems though. Because if one of the ideal systems we are comparing is unstable or comparatively less stable (with extremely bad outcomes), we would likely favor the less quote-unquote ideal system
2023-09-08T05:16:15+00:00 | 🔗
@timschlomi Both, but I think "can" is a more of a tamer line of reasoning of just practical difficulty
2023-09-08T05:15:00+00:00 | 🔗
@timschlomi 2. But then can these predicting agents actually account in each individual case, efficiently at scale, when it seems to me that even the best-in-class, aren't extremely confident when it comes to their assessment of the trial outcomes. Even when they are looking at the jury
2023-09-08T05:12:11+00:00 | 🔗
@timschlomi I see ur point here, but ig practically if you are going after that then its the Constitution and so practically speaking the proposal is dead on arrival
2023-09-08T05:08:52+00:00 | 🔗
@timschlomi I guess my claim then is that if all this is the case and all of the various nuances are resolved, and we trigger these trials, then I don't think its obvious that we will have different outcomes (in terms of number of trials, pleas etc). especially given the cost.
2023-09-08T05:06:44+00:00 | 🔗
@timschlomi I think that that uncertainty there is so high and the bad outcomes are so bad that practically it isn't feasible