I noticed this because I set my homepage to the AI Countdown clock, which is based on this question. It's deeply unnerving.

Seems to me like we might be underestimating the probability that an influx of new users will drive it up some.

Would anyone who codes be interested in doing a Metaculus-centric Hackathon sometime? "Met-hack-ulus"?

As more outbreaks become self-sustaining and begin their exponential growth phase, there will be fewer cases with any link to China. I suspect that this number will decline rapidly enough that WHO will drop it from the sitreps within a few days. The only question is thus what will it be by the time the drop is made?

Almost literally every time I see some news relevant to a question, I copy the link and rush over to Metaculus, only to find that @Jgalt has already posted it. You are incredible!

@kokotajlod Write it up and see if metaculus agrees!

@(Tamay) By 2020-01-20, there were [282 cases and 6 deaths reported](https://www.who.int/docs/default-source/coronaviruse/situation-reports/20200121-sitrep-1-2019-ncov.pdf) by the WHO. However, according to the [*Clinical Features*](http://dx.doi.org/10.1016/S0140-6736(20)30183-5) paper, those 6 must have been a part of the original 41 infections analyzed throughout December, implying a mortality rate of of ~15%. Because of this, I think we should launch the question on the number of deaths, or clone the modified version of this question with the scale r...

@Sylvain Let's consider some reasons why it might not be desirable:

  1. It could increase the onus of question construction, resulting in a reduction in the number of questions written.
  2. It could cause questions to become somewhat more formulaic, leaving less opportunity for questions with creative structures (e.g. How Impeached Will Trump Be?)

Neither of these strikes me as sufficiently compelling, important, or likely to be dissuasive.

Metaculus Swag Shop. In particular, would like:

  1. T-shirt with logo and a QR code on the back linking to https://metaculus.com/questions
  2. Hoodie
  3. Coffee Mug
  4. Laptop stickers
  5. Autographed copies of Cosmological Koans...

The value of the clock is how threateningly close it feels to midnight. If we actually get there, it's sort of like giving up. The marketing value of it would immediately dissolve. Accordingly, I predict that the BAS will never do it, and this question will be resolved ambiguously.

There have been [614 papers in the first 56 days of 2020](https://scholar.google.com/scholar?q=%22Climate+engineering%22+OR+%22geoengineering%22+OR+%22Solar+radiation+management%22+OR+%22Stratospheric+Particle+Injection%22+OR+%22Stratospheric+aerosol+injection%22+OR+%22Marine+cloud+brightening%22+OR+%22Ocean+sulfur+cycle+enhancement%22&hl=en&as_sdt=0%2C5&as_vis=1&as_ylo=2020&as_yhi=2020). Assuming this proportion holds, we should expect ~4000 papers. But I don't think this is as strong a basis for prediction as linear extrapolation from prior years, whic...

@DanielFilan IMO, one of Metaculus' greatest weaknesses is the lack of short-fuse questions, making it very unrewarding for newbies. Election questions have the advantage that they don't require very long term investments of attention to pay off.

Elon Musk Prediction Heuristic: take his timeline, and double it. If he suggests in 2016 that a mission to Mars with humans will launch in 2024, look for that mission to actually occur around 2032. There will be 5 oppositions before this question is resolved, of which the latter three are actually the least optimal for missions (purely in terms of distance). So there are a few good reasons to think it might not work out within Musk's aggressive timeline.

It has been 54 days since this question posted, during which time the number has dropped from 836 to 833, implying a background rate of -3/54. Assuming this is just "business-as-usual", we should expect around 20 fewer employees at resolution time, or roughly 806. It appears ordinary employee traffic is highly unlikely to trigger the positive resolution.

@Cory Improved bookmarklet a lot. On binary questions it should work in a pretty intuitive way. For range questions, it outputs the 0-1 normalized mean, spread, and weight for each component. (Still figuring out how those get translated back to the range.)