Uncertainty about base rates of NYT/etc. reporting prescience
No Scores Yet
Relative Brier Score
6
Forecasts
3
Upvotes
Forecasting Calendar
Past Week | Past Month | Past Year | This Season | All Time | |
---|---|---|---|---|---|
Forecasts | 0 | 0 | 6 | 6 | 60 |
Comments | 0 | 0 | 5 | 5 | 23 |
Questions Forecasted | 0 | 0 | 2 | 2 | 19 |
Upvotes on Comments By This User | 0 | 0 | 3 | 3 | 13 |
Definitions |
Star Commenter - Feb 2024
I don't really have much signal beyond the crowd and a very noisy line of reasoning that closing down seems out of the ordinary + excessive in the face of alternatives (e.g., scaling back operations, relocating top researchers).
Perhaps worth considering that the US and China are supposedly trying to have dialogue on AI safety/regulations, so unless those talks dramatically burn out, I don't think the US would like for Microsoft to do this.
I would love to have more base rates on companies/Microsoft closing research labs in China.
The fact that the closure doesn't have to occur by Dec 31, 2024 seems noteworthy, otherwise there would be less time to actually enact the closure.
Now that I think about it, perhaps another good base rate here would be "how often do top-level media reports about company debates precede such drastic actions?"
Like, maybe it's really worth something that NYT is reporting about it?
Interestingly, only one exoplanet has reportedly been discovered by the James Webb Telescope: https://www.hpcf.upr.edu/~abel/phl/hwc/data/hwc.csv (WISE J033605.05-014350.4 b, discovered on 2023-05-03).
Indeed - and it is not a habitable one. Nice catch!
UPDATE: seems we have 2 more candidate ones as of Jan 2024, although still not habitable (giant planets): https://arxiv.org/abs/2401.13153.
Adjusting downwards towards base rates of roughly 73% ( 8 times in the past 11 years, it seems: https://www.hpcf.upr.edu/~abel/phl/hwc/data/hwc.csv ). Above 73% because last two years were 5--right on the edge--but technological improvement doesn't seem to have produced a large number of new discoveries... It might be right on the line, like 2022 and 2023.
Why do you think you're right?
Base rates
Why might you be wrong?
Spent almost no time thinking beyond base rates.
Technically, I originally just meant "the fact that it happened last year, and seemed like it happens moderately frequently."
However, it turns out the base rate over the past decade seems to be around 80%, per data from https://phl.upr.edu/hwc/data (specifically, https://www.hpcf.upr.edu/~abel/phl/hwc/data/hwc.csv ). Starting in 2013, the years 2013, 2014, 2015, 2016, 2017, 2020, 2022, 2023 had at least 5 habitable planets. That's ~72%. Seeing this data in more detail, I might adjust downwards a bit, back to 80%. (It's concerning that the gaps were more recent -- 2018, 2019, 2021 -- but it's unclear why this is the case.)
Updating based on crowd.
Updating down based on analysis here: https://www.infer-pub.com/comments/120891
(H/T ctsats)