Science/math stuff

  • Experimental confirmation of the Many-Worlds interpretation

  • Definitive evidence of intelligent extraterrestrial life

  • Discovery of Earth-based life unrelated to known life (different chemistry, unrelated genetic code, etc)

  • Evidence that life on Earth originated elsewhere (panspermia, etc).

  • Proof that P=NP

  • Disconfirmation of the Riemann hypothesis

I would have a much more confident answer to this question if I knew whether Gogo InFlight (InSpace) Wi-Fi would be involved.

If yes, I would answer negatively, although I would answer a question affirmatively that asked "Will a Metaculus user repeatedly try to post a comment on this question from space, and fail, before 2050?" Thank you.

I've made this request before, but re-upping it: Please make a much stronger visual differentiation in the question page between questions I have made predictions on vs. questions I haven't. This seems especially bad for some reason in the case of numerical predictions. Case in point: I clicked through on a question that asked about the number of coronavirus infections, and included "(Question two)" in the title. I had answered a similar question before, and missed the "Question two" bit, and so stared stupidly at the question entry display, which said...

Please set up appropriate AIs triggered from cron jobs to judge question completion and award points. The fact that no human may be left alive does not absolve Metaculus-the-system from doing the right thing in the positive case.

I like the idea of bundling these by topic. E.G. a set of unlikely political events, a set of unlikely scientific discoveries, etc.

(I was tempted to reply to Anthony's [Metaculus capabilities list]( with this request, but this is a better place for it.) I'd like to renew my previous suggestion of category-specific leaderboards. I'm realizing that there are two different ways you could approach it as a UX matter. You could 1) Have 'L' links next to each eligible category on the Categories page, with the links being to that category's leaderboard. Or you could 2) Have categories as a filter on the Ranking page. I...
This was a very interesting trio of questions for me to try to answer. Because if we have D = how many people will die from infection in the timeframe I = how many people will have the infection in the timeframe, and R = the ratio of fatalities to infections if I answer the three questions independently and naively - guesses off the top of my head - I get values of D, I, and R such that D << I * R which I know can't all be right. I think that this is because a particular value of I can seem entirely reasonable to me, and same with a particular ...

I still find it a little confusing that a question I have not yet answered is so visually similar to questions I have answered.

One very simple cosmetic change you could make for binary questions would be to replace the "you: 50%" label on the slider control with "you: ??%" when you have not yet made a prediction.

@(Uncle Jeff) wrote: > What this is really asking is whether the sponsor behind the organizers will decide to toss another pot of money at buying a bigger protest. If this were the US, the sponsor would be George Soros, and we'd want to know what his endgame was (perhaps by paying attention to one of his mouthpiece front orgs). I don't know who the puppet-masters are in France, so I'm staying out of this one. I'm impressed at the skill with which you lob a few gratuitous political grenades before you announce that you're "staying out of this one". :) (Y...

Wow, having played with this now I am really impressed by this feature. The UI is so intuitive with multiple components, and yet it imposes no extra complexity if you just want one component - there's just one extra button to avoid clicking on. Bravo.

I vote for more social visibility into our predictors and commenters. In addition to celebrating people who are expert in a particular category, I would like links from a user's profile to authored questions, and also comments by the user directly in line in their profile. Presentation of the comments might be challenging, since they would lack context, but I think the comment text plus a link to the question would be OK. Better yet would be if the link could be to an anchor in the question comment stream itself (in addition to the text inline in the pr...
This must have been suggested many times, but I'd love to see multiple-choice questions where you assign a probability to each option, summing to 100% (including a none-of-the-above case). This could cash out to individual yes-or-no predictions under the hood, but would be easier for users to track, and also could enforce the 100% sum. For example, I like the question about whether Elizabeth Warren will be the 2020 Democratic nominee, and would like to see analogous questions for all the other potential candidates. Seems like a headache to have all the...
My prediction here is 1%, so I am convinced that the combinatorics are completely prohibitive of a solution. But it's interesting to note that having a deep game tree is not equivalent to being hard to solve. As a fun counter-example, take this modification of chess: Embed a 8x8 chessboard in the middle of a 10x10 chessboard, so that there's an outer strip of empty squares at the start of the game. Keep all the rules the same except that: 1) kings may not move to the outer strip, and 2) if any other piece moves to the outer strip, then that player wi...

Congrats Gaia - great to see this happen.

2042 sounds early. They could be really behind on their email. Let's not assume that we'll be at the top of the queue.

Although factors mentioned below are likely to affect the log score going forward, I think the most important factor is difficult to model predictively: how hard will future Metaculus questions be? This will depend on makeup of the Metaculus site owners, makeup of the question-submitting user base, and fashions for styles of question.

If I understand the scoring correctly, all we would need for the log score to drop dramatically would be a raft of questions like "Will the Sun rise tomorrow?"

The tournament registration choices for academic field of study includes

  • Computer
  • Science

as separate successive entries. Looks like a linebreak snuck in somewhere.

I think this is a great question, and I have no suggestions to improve it. Just for discussion, though, I want to point out that it the question implicitly identifies human intelligence with abstract reasoning and "book-learning", rather than embodied, perceptually-driven, commonsense reasoning and action. This is a similar bias to the one that led early AI researchers to assume that vision would be easy and chess would be hard. Here's a personal-assistant-task test I'd like to give to a robot (which maybe could form the basis of a different question)...

Very exciting - welcome Joanna, Dan, and Juan