From 5412dd290480dd0e3068afe3f7022f19ec39ec42 Mon Sep 17 00:00:00 2001 From: Nuno Sempere Date: Tue, 29 Sep 2020 17:17:56 +0200 Subject: [PATCH] Updated list of projects --- ea/Forecasting/ListOfForecastingProjects.md | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-) diff --git a/ea/Forecasting/ListOfForecastingProjects.md b/ea/Forecasting/ListOfForecastingProjects.md index cf17dc7..7a26a9a 100644 --- a/ea/Forecasting/ListOfForecastingProjects.md +++ b/ea/Forecasting/ListOfForecastingProjects.md @@ -1,14 +1,14 @@ # List of forecasting projects: - Have forecasters read Toby Ord's book, and generate their own probabilities of x-risks -- Feed forecasted x-risk probabilities, in addition to some other inputs which could also be forecasted, to Trammel's model on patient philantropists, to find out what the optimal spending conditions for a patient philantropist would be. Right now, some of the inputs to Trammel's model are very much up in the air. -- Forecast whether some big companies will keep to their chicken broiler / animal suffering commitments; this would be useful for some orgs to prioritize their outreach. -- Forecast the value of altruistic projects; I (Nuño) have a small demo on this. +- Forecast variable inputs to Trammell's model +- Forecast whether some big companies will keep to their chicken broiler / animal suffering commitments +- Forecast the value of altruistic projects; I (Nuño) have a small demo on this - Throw some optimization power at the Metaculus Ragnarok questions - Look into the relationship between price or time spent and forecast accuracy - Look into rewarding forecasters according to their Shapley values - Predict which EA Fund applications will be funded -- Add interesting and nontrivial markets to Augur. +- Add interesting and nontrivial markets to Augur - Rapid Response EA Forecasting Hotline - Forecast incubator ideas' likelihood of success - Create a calibration hall of fame/shame for researchers' past forecasts @@ -16,9 +16,9 @@ - Pay and support forecasters to make forecasts on existing platforms - Get a visible forecasting win for EA/QURI/EpiFor - Participate in an OSINT project -- Suggest questions to Good Judgement Open with an altruistic bent. It is not unusual for GJOpen to accept question suggestions. -- Talk with CSET; they seem a little bit confused/lost. -- Recruit people for CSET; they need more forecasters to be able to produce useful policy feedback +- Suggest questions to Good Judgement Open with an altruistic bent +- Talk with CSET; they seem a little bit confused/lost +- Recruit people for CSET - Forecasting for EA policy advocates - Estimate the value of the forecasted variables/questions on Metaculus and GJP - On-demand forecasting to research teams