I was surprised that GPT-3 was able to answer this correctly from a transcript, and I wondered if maybe discussions of the book could have been included in its pretraining, facilitating the answer. However, some simple variations of the transcript were answered well. One of the most “difficult” variations I wrote is [this one](https://pastebin.com/ciq8UtV7) (everything after the first “Answer: ” is the first completion of GPT-3). In the end, I wondered how much of the answer was based on the information from the text, and how much was following from t...

I only got 1 point from this resolution, having a final prediction at 99%. Seems a bit low.

I could not find any confirmation that these cows were actually infected with bovine bluetongue, but assuming these culling events count for the purpose of this question, we have a lower bound of 850+1610.

From [First return then explore](https://arxiv.org/pdf/2004.12919v3.pdf) >On Montezuma’s Revenge, the mean score of the robustified policies is 1,731,645, exceeding the previous state of the art score by a factor of 150. This post-robustification score on Montezuma’s Revenge is not only much greater than average human performance, but in fact is greater than the human world record of 1.2 million. Notably, the performance on Montezuma’s Revenge is limited not by the quality of the learned agents, but by the default time limit of 400,000 frames that is i...
It seems like the O*NET data is mostly based on [standardized questionnaires from workers and occupation experts, plus ratings from occupation analysts](https://onet.rti.org/survey.cfm). In particular, it appears like the degree of automation is mostly derived from [question 49 of the following questionnaire](https://www.onetcenter.org/dl_files/MS_Word/Work_Context.pdf). From what I can tell, there is an apparent absence of objective measures or criteria, so it is now unclear to me how an increasing automation would actually be reflected in the "degree...
*"When will an algorithm be able to predict the [Big Five personality traits](https://en.wikipedia.org/wiki/Big_Five_personality_traits) of a person from a naturalistic photograph or video?"* I think this would have far-reaching implications, and might not be too far off given results such as [this](https://www.nature.com/articles/s41598-020-79310-1). I'm unsure on how to precisely formulate the question and its resolution criteria. I was thinking that it could resolve positively if a trustworthy publication claims that an algorithm can do this on som...

I wish Metaculus had a better system for assigning probabilities in discrete non-binary questions. I can imagine some (low-probability) proliferation scenarios where nuclear weapons disseminate to many countries, but currently I have not enough bins to represent that without "losing" a lot of probability mass to non-integer values.

@ugandamaximum One possibility that comes to mind is that they model a portion of the population becoming less willing to vaccinate if booster doses become recommended, as such recommendation would somehow imply a lesser effectivity per dose and more hassle to become “fully vaccinated”.

@MichaelA @admins I really think this question should be separated into two different ones: A binary "Any nuclear weapons used by 2050" and a conditional "Number of nuclear weapons used by 2050, if >0". It seems like precise forecasting on this subject is very important, but the current formulation loses a lot of information because of the frontier between 0 and 1 and the relative high probability of zero.

Given the distant closing date, would it be possible to resolve this question as ambiguous and open these two others?

Currently, the fully vaccinated percentage for the age group 12-15 is of 48.6%. It seems like there was some big update recently.

I like this change.

From what I understand, the new system will also apply to closed but unresolved questions. Would it be possible to avoid applying the new scoring method to old contests that relied on points? I'm thinking in particular of the Deep Learning round of the AI progress tournament, whose scoring also interacts with the Bayes Prize from the same tournament.

Can somebody explain why Chernobyl sits at the same level as Fukushima in this logarithmic scale? The former seems to have been an order of magnitude worse in terms of radioactivity released, and it also seems much worse in fatalities (both short term and long term).

I think that questions that are conditional on some event happening are particularly useful, both for policy considerations and because the prediction becomes more legible (since the same question, without the conditional, combines the probability distributions of the negative and positive cases). A particular suggestion would be the following: If a COVID-19 vaccine becomes available for children under 12 during October, how many children under 12 will by vaccinated by the end of 2021? Same questions for November or December might be useful, maybe mo...

Sadly for my points, it seems like this resolves as 23.2. @admins

This comment was moved from What will the average top price performance (in G3D Mark /$) of the best available GPU be, on Jan 14, 2022? to this question group. Learn more about Question groups

The question is restricted to the approval of products for human consumption, right?

@RyanBeck I think so. At the end of the document it says that it terminates Seventh Amended Number Seventy-Two (2021) and Order of Public Health Emergency Nine, which includes the point I.A.2, whose subpoints b and g state the 6-foot distance restriction.

@Rexracer63 Note that the dashboard is currently showing only the numbers for confirmed hospitalizations. I think we should wait until we have a source that includes the pending hospitalizations.