@Jotto I am confident at 1% on this question, and would go lower if metaculus supported it. I am a little less confident on the sister question about landing anything on Mars. [1] There I am at 4%, greatly down from the question opening based on Musk dropping the original Red Dragon plan and no new details being forthcoming.

[1] https://www.metaculus.com/questions/737/will-…

@Linch I think we are at 4 out of 100 senators who have had a positive COVID test. You really think Trump's risk ratio is over 10 times as high as a typical senator? My take on this is the metaculus prediction was better than the community prediction, and it is basically chance that the resolution is positive despite the low prior odds of the event.

OpenAI has a new paper out on their Dota agent. See https://cdn.openai.com/dota-2.pdf. It reports compute usage of 770 Petaflops/s-days which I calculate as 6.6e22 Flops. So about a factor of 10 below positive resolution.

You guys are nuts! Major parts that have to be built before landing people on Mars: (1) Earth launch vehicle (2) interplanetary spaceship (3) Mars lander (4) Mars surface kit (airlocks, spacesuits, etc) (5) Mars launch vehicle (6) Earth return capsule That is assuming an Apollo-style mission with command/service/lander modules and a minimal Mars stay. Other mission designs would need even larger engineering components. If the Dragon can be re-used as an Earth return capsule, then that is the only element above with hardware under test t...

Some names sorted by birth year, divided by generations:

1941 : Bernie Sanders

1942 : Joe Biden


1947 : Hillary Clinton

1949 : Elizabeth Warren

1952 : Sherrod Brown

1954 : Oprah Winfrey

1956 : Deval Patrick

1957 : Andrew Cuomo

1957 : Terry McAuliffe

1964 : Kamala Harris


1966 : Kirsten Gillibrand

1969 : Cory Booker

1973 : Chris Murphy

— edited by traviswfisher

Some thoughts, in random order: * Give community prediction and metaculus prediction points and show their position on leaderboards. Award badges and/or points to users who come in ahead of those benchmarks. E.g. the "better than community prediction score in last 30 days" badge, which can be earned many times, perhaps once per day or once per 30 days. (Rationale : paying power users in ego boosts encourages better predictions.) * Go back to adding immediate points for the earliest predictors of questions. (Rationale : encourage users to check the ...

@Sylvain @v This dates back to v's time dominating the metaculus prediction level. Kudos on the foresight, v.

Note the resolution condition here is based on JHU or WHO data, both of which are baaed on confirmed cases by testing. But the US has been unable to scale up testing and is now giving up on doing so [1]. Other countries are also unlikely to keep testing up in the kind of mass infection scenario that would give a posotove resolution. There is a significant likelihood the disease is this deadly but this question still resolves negative.

[1] https://www.washingtonpost.com/health/2020/03…

— edited by traviswfisher

One question mark is whether the resolution source (COVID-tracking spreadsheet) might switch methodology as other sites have. The CDC updated guidelines on April 14 to count "probable" cases, which lead to NYC revising death tolls upwards by over 3000 people. Other locations are also releasing updated, higher death toll numbers. The COVID tracking project describes itself as: "The COVID Tracking Project collects and publishes the most complete testing data available for US states and territories." This focus on testing probably means the methodology ...

The deep mind team is sitting on hardware that can train alphagozero in 3 days. Presumably they will find SOMETHING to do with that hardware over the rest of this year.

The bigger question is, not will it be done, but will it be published?

The Olympic Committee is saying the go/no-go decision date will be towards the end of May.


@Jgalt From what I am seeing I am afraid that the right target is estimated percent of world population infected. If the latest estimate of R0 ~ 2.5 is at all close to reality, this one is headed to full pandemic territory.

Related, unforthnately probably short-term question: will WHO declare this to be a pandemic?

I think this resolved incorrectly. The "fine print" says """This question resolves as the date when the first credible article is published that presents the claim that at least 10 million people have been administered a vaccine that had been previously shown to have an efficacy rate exceeding >75% in a randomized controlled trial with at least 500 participants.""" I don't see how that is compatible with resolution to a future date. There should be a qualifying news article linked for resolution, and it should be as of the date of that article. BTW...

<1% for either to have a permanent population by 2100, much less of thousands. It's silly to speculate between two not-gonna-happens.

@notany I'm not sure if I'm off the hype rail or not. After some intensive self-deliberation I am at 93% on this question currently, down from 99% prior to the emergence of this virus.

Is it required that the language about "reduced neutralization by convalescent and post-vaccination sera" be posted on the CDC site?

Here for instance is a preprint saying "Highlights: Original strain convalescent and vaccine sera show reduced B.1.1.7 neutralization": https://www.sciencedirect.com/science/article…

For what its worth, Boston's ban passed Wednesday, June 24 so I think that one should be the reference for question close date. https://www.google.com/amp/s/amp.wbur.org/new…

For what its worth, I put a prediction of 74% on January 22 and didn't update it. I agree that it could have gone higher by question close time. The metaculus prediction skews lower than the average prediction, probably incorrectly in this case, but statistically we humans are proven to be over-optimistic in predicting.

@dan I think this resolved too early. NYC has done a data update; now every day before May 6 shows over 200 hospitalizations.