I dislike this type of "vs." questions. Binding two non-related questions together without good reason:

  1. provides no additional insight,
  2. increases the uncertainty and variance compared to two separate questions.

@traviswfisher Very unlikely. World population grows over 80 million per year.

It has to be Spanish flu level pandemic to reduce world population temporarily. Spanish flu was so deadly partly because WWII created large number of injured and malnourished people. Today medicine, containment etc. are much better so even Spanish flu could not be so fatal.

A plot of metaculus users.

X-axis: number of resolved predictions by user.
Y-axis: brier score or some other proper score function (I would prefer logarithmic rule)

It would be nice to have interactive plot where you can see the username of the point in the plot.

— edited by notany

It's possible that current model like GPT-3 could do it just based on text. Not because they can figure it out from the clues, but because "Bear eats X" is a typical thing for a bear to do in stories. If the model recognizes the generic tone and narrative style it can often give a correct answer without thinking. Children do this too. You give them list of math questions, then finish with something like: "John has 50 apples. John's father ate 40 apples. How old is John?" Many children just do 50-40 = 10 because 10 is a good age for a boy with a fathe...
Update to my earlier prediction: https://www.metaculus.com/questions/2671/will-the-united-states-fulfill-its-goal-of-landing-humans-on-the-moon-again-before-2025/#comment-13030 Still 1% with more evidence. NASA's own audit does not believe the timeline either. Covid has delayed the progress even more than anticipated. Report November 12, 2020 NASAS, Office of Inspector General https://oig.nasa.gov/docs/MC-2020.pdf >Challenge 1: Landing the First Woman and the Next Man on the Moon by 2024 >... Due to these challenges, we anticipate further schedule de...

It seems that the current plan is to get first light 2025, so it's very likely that the answer is no.

@gjm I have experienced the same in many other forums.

Signal to noise ratio simply declines if the forum is free for all and becomes popular, see September that never ended.

You have to form a culture with moderation, boundaries and defenses to keep the quality high. It's not automatic. Posting guidelines and rules on the side is jut not enough.

This is good question for multi-modal prediction.

I suspect that either just 0-3 Republican senators vote to convict, or 18-25 senators vote to convict. Senators don't put their neck out against Trump if it looks like he is acquitted and re-elected. But if it looks like he would lose, or there are enough votes to convict more senators will vote against.

Einstein field equations allows many weird solutions when you just play with them mathematically and ignore physics or invent your own. Think some weird geometry and then invent nonphysical matter or matter configuration that provides the solution. The only reason I can see why Alcubierre drive being so interesting compared to other solutions is the name "warp-time" and "drive". It invokes the sense that it's a mechanical device you can build and not just a mathematical exercise. If it was just called Alcubierre solution, It would not be any more int...
Dementia incidence increases exponentially with age between the ages of 65 and 90 years and doubles approximately every 5 years. If I remember correctly, male in Biden's age might have 1-2% per year change of getting dementia. When it comes to visible symptoms, I don't think there is reasonable evidence. 1. Biden has always been susceptible for gaffes. It's called *"Joe being Joe"*. When he was younger it was just the way he was. Now it's seen as a symptom. 2. healthy 77-year-old without dementia can have problems with recall and short episodes o...
I estimate that the minimal computational power necessary to exceed human general intelligence with requires only the raw computational power of very small mammal or a bird, like tiny mice or hummingbird (there are researchers who agree with me but I'm not saying names because I don't remember who they are and I'm too lazy for literature search). The caveat is that we may not be able to have algorithms that get even close to minimum. We may need million times the computational capacity of the human brain to replicate human intelligence. Just like th...

@jahooma Not a big deal. Iowa is only 6 electoral votes.

Most forecasts (Including The Economist forecast that gives Trump only 4% change) have predicted that Trump wins Iowa since the beginning.

Africa has zero cases according to Johns Hopkins data.

It may be because only Senegal and South Africa can currently diagnose coronavirus. Most of them should have received the ability as of Friday, so they may be more cases soon. There are over 200,000 Chinese workers in Africa.

— edited by notany

It just occurred to me that probability of war is endogenous variable in the nuclear brinkmanship game.

If the the perceived probability is too small, it will not work as a deterrent or threat. Participants must increase the probability (accidental escalation or perceived irrationality of player) until the threat becomes credible again. The probability of nuclear war has probably some equilibrium.

— edited by notany

@Jgalt It's very straightforward calculation based on herd immunity threshold. If R0 is between 2.5 and 3.3 the proportion of infected will be between 60% and 70% when the epidemic dies out.

This assumes that the spreading and effective reproduction rate can't be controlled using other means.

— edited by notany

Quick and dirty estimate: 1. (p=0.3) From purely technological and infrastructure perspective feasible (still high risk of failure) in the near future assuming no economic or resource constraints. Project can draw unlimited funds and resources. 1. (p=0.3) Physiological feasibility unknown. If people confine themselves living under several meters of regolith, radiation exposure can be managed. The long term effects of 0.38g gravity are unknown. See, [Death on Mars - The martian radiation environment is a problem for human explorers that cannot be overst...
[Law Firm Stops Representing Trump Campaign in Pennsylvania Suit](https://www.nytimes.com/2020/11/13/business/porter-wright-trump-pennsylvania.html) "Porter Wright said in a court filing that it would no longer represent the campaign in a federal lawsuit alleging widespread voter irregularities." Related to earlier report: [Growing Discomfort at Law Firms Representing Trump in Election Lawsuits](https://www.nytimes.com/2020/11/09/business/jones-day-trump-election-lawsuits.html?) Lawyers can honorably represent reprehensible clients and causes. Law fir...
The questions seem "made up" or forced. They have little or nothing to do with AI Progress. Alphabet Inc. is an advertising company. Their market cap six months from now or Industrial activity six months from now does not measure AI progress. It's the question of economic cycles, international trade, antitrust issues, etc. The value of Object Detection and Image Classification Indexes in the short term depends more on conference schedules. In the longer term, it depends on how many models are tested on these older benchmarks. Hard pass on this compet...

Those 1 TB SD cards are for special purpose use. You can't compare them to general purpose HDD's.

SandDisk card memory lasts only 512 writes and it's intended for serial writing. If you record surveillance video 60 MB/s continuously and overwrite only after the card is full, the card lasts 16 years and can store last 11 days.

@AngraMainyu Elon Musk's years aka Mars years, 1.88 Earth years.

"SpaceX will land humans on Mars about 6 years from now."

6×1.88 = 11.28 Earth years (year 2032).