I think it has to be the all-things-considered view, or else Metaculus becomes one of those games that ethically require their players to play badly in such a way that they constantly have to think about exactly how badly they're ethically required to play. I'd be in favor of eliciting inside view predictions separately.

— edited by steven0461

@Jgalt "When will the Bulletin of the Atomic Scientists announce we've reached 1am?"

Experimental question idea: "Conditional on EITHER a randomly generated number between 1 and 100 coming up as 1, OR a nuclear war happening in 2022, or both, will a nuclear war happen in 2022?"

The benefit of asking it this way is you get meaningful resolution, e.g. if you think the probability of a nuclear war in 2022 is 0.2%, you'll want to predict ~1/6 or 17%. The cost is the question will most likely have to resolve ambiguous.

Zeke Hausfather on Twitter:

2022 is unlikely to set a new temperature record if these forecasts pan out (though uncertainties are too large to be too confident at this point).

There's one more month of data since that tweet, but it doesn't look to me like it would change the conclusion much.

Metaculus currently assigns a 37% probability that under 2% of the world population will be infected, so I think that's still/again not consistent with the 93% estimate here.

https://www.gwern.net/Embryo-selection > genome synthesis is a similar level of seriousness, but is much more predictable and can be looked for, very loosely, 2030-2040 (and possibly sooner) https://www.gwern.net/Forking-Paths > IES has a considerable headstart but genome synthesis progresses like a metronome because if it gets blocked on one track, it can hop to another; so my money is on genome synthesis at the moment. Searching Metaculus, I could only find references to genome synthesis for viruses. I would be interested to see questions about gen...

Sea ice decline slowed down a lot in the last few weeks and it now looks virtually certain that this will resolve negatively.

@Roko Is it really a positive transition to superintelligence if we still need to use Metaculus to judge each other's smarts?

— edited by steven0461

This seems just a straightforward matter of almost-linearly extrapolating from the past years to something like 415 +/- 1. I'm guessing based on the text that there will be more questions like this for later periods, when uncertainty about emissions becomes more relevant.

If my interpretation is correct, Carbon Brief is estimating 100%. (2018 was fourth warmest, so 2019 would have to be fifth warmest for a negative resolution.)

— edited by steven0461


In this paper, we ask: which scenarios--both baseline and policy--have fossil-fuel CO2 emissions growth rates most consistent with observations + IEA 2020 Stated Policies from 2005-2040, and what do these scenarios project, looking to 2100?

Answer: These scenarios project about 2-2.5 degrees C of warming by 2100.

@(JonathanShi) Some points in response to the shy voters link: 1. Underestimation of Trump's vote share in polls being correlated with the number of Trump voters in the state seems consistent with a model where some fixed fraction of Trump voters say that they won't vote for Trump. 2. The direct estimates of the fraction of shy voters given are 2% (in October) and 3% (a few months in advance) decreasing to 1-2% (just before election day). It's now a few months in advance, which suggests the 3% number is the most relevant. I don't understand the 538 mod...

It doesn't fit the stated criteria, but I'd be very interested in a question like "Will the 2020 election have a weird outcome?", resolving positive under any of the following conditions and probably others I haven't thought of:

  • the winner is someone who, as of 2019, is not primarily known as a politician
  • the winner is not a Democrat or Republican
  • the outcome is seriously disputed
  • the winner of the election does not end up in power
@(lowiqwizardry) I don't think the Fermi paradox is informative here. If I learned the question would resolve positively, it would be a small update about how likely current human circumstances (including beliefs and intentions) are to lead to space colonization. It would be a much smaller update still about how likely random alien circumstances are to lead to space colonization. For zero stars to have spawned a noticeable civilization, you need many orders of magnitude of improbability per star, and considerations like this only get you a tiny part of ...
What if, out of every 1000 civilizations, 999 stop being detectable after one century and 1 colonizes the universe and remains detectable for trillions of years? Taking the question literally, I don't see a plausible case for anything below the top end of the answer range, unless maybe it involves extremely convergent incentives to leave no trace or a relatively early end to the whole universe. (edit: it takes less than trillions of years for most galaxies to become unreachable, so there's that too; basically I think the question assumes a bad model and ...


Looking ahead, Vose said, there's a 99 per cent chance that 2022 will rank among the Top 10 warmest, a 50/50 chance or less that it will rank in the Top 5, and a 10 per cent chance that it will rank first.

@joshua_gem Spikes at the end of the interval are pretty common. It seems to be how the site displays out-of-interval probability for some reason - 12% in this case, which seems low to me. I don't think people put their actual peak at 2200.

[Zeke Hausfather on Twitter](https://twitter.com/hausfath/status/1247196142955184128): > At this point 2020 still looks like it will give 2016 a run for its money as the warmest year on record (though I'd personally bet on 2020 more likely than not being the second warmest) I don't understand the "more likely than not" given how narrow the window between first warmest and third warmest is, but this seems like an update toward expecting 2020 to be warmer than 2019. As Hausfather points out elsewhere, COVID-19 has little effect here because almost all gr...