@(alexanderpolta) wrote: > I also don't believe for a second that it's a coincidence that the "HIV cure" just happens to be a known genetic-edit to increase intelligence. > > It's the perfect rationalization to try, and it affects the thing Chinese officials are most likely aiming to select for in their burgeoning eugenics programs. This isn't a robust result at all. Existing literature tying low or blocked CCR5 to any kind of increased mental performance is mainly regarding how it might improve recovery from stroke. There is also one mouse study from ...
Did one of those fancy Google Sheets that the real-ass Metaculators be making. Warning: it describes all the main character deaths in Game of Thrones and thus involves massive spoilers. Using the top table here as a list of "main characters": https://en.wikipedia.org/wiki/List_of_Game_of_Thrones_characters Sheet: https://docs.google.com/spreadsheets/d/1NDJBlnH-Q9URJVpzGMGLYSoE0kWXCxNWjxq3aKR9ePw/edit?usp=sharing (may have missed some deaths) Turns out that the mean season of Game of Thrones only kills about 8% of its remaining main characters. edit...
Assigning ≤ 1% to the summed probabilities of the following (accounting for necessary time-frames and unwritten requisites): - Musk dying from any cause - Musk becoming sufficiently ill from any cause that he can't go - Musk heeding advice saying that it's not a good idea for him to go - Musk secretly being too scared to want to go despite public statements - Musk changing his mind for some other reason - Musk trying but not surviving "to a height of at least 100 km" (resolution criteria) - Musk deciding to try for a different celestial body instead - S...
@(Linch) I also think that the Metaculus prediction for continuous questions could be improved. I think it's pretty likely that the algorithm underlying it could be improved, and further, I think it's somewhat likely that the UI for making continuous predictions impairs individual predictors' abilities to make their best predictions. Regarding the algorithm: The Metaculus prediction must be doing *something* like weighting (and transforming?) predictions from individuals according to their previous performance. The "yellow" Metaculus prediction for cont...

Here's my model. I might be doing some of the maths wrong, and some of the many parameters fed into it might be iffy. It basically expresses this idea: natural epidemics are quite common, and epidemics seeded from lab-acquired infections are much rarer, so you need strong evidence to adjust your posterior to anything much higher than a single-digit percentage chance.

@kimbob14 wrote:

If he continues to keep his promises, I believe he'll be re-elected. I pray for him every day and so far God has answered my prayers. GBU

HEll yeah dude

Of the four previous Ebola outbreaks that reached 200 deaths, one of them (the most recent) resulted in a death toll of > 400.

The number of deaths is still increasing (e.g. 9 deaths between the 9th and 16th of this month). There is no obvious increase or decrease in the gradient of deaths over time across the past month.

@KrisMoore Update: I've now had something like 10 PCR tests and an antibody test, all negative.

Wikipedia list of notable Twitter suspensions

I've scanned the list and can only spot two national-level, then-serving politicians who've been suspended: Australian Senator Pauline Hanson and Israeli Knesset member Bezalel Smotrich.

— edited by KrisMoore

Note that among Chinese patients [COVID-19 mortality rate was apparently raised by certain pre-existing conditions]( https://ourworldindata.org/coronavirus#case-fatality-rate-of-covid-19-by-preexisting-health-conditions), notably hypertension + cardiovascular disease. (Somebody more motivated than me could dig into the primary data and see if there's a way of getting mortality by both pre-existing condition + age.) [Biden's blood pressure is high-ish](https://go.joebiden.com/page/-/CiU2iAa4Ig_%20j5EQzueoad.pdf) but not clinically so. Bernie obviously had...
@(KrisMoore) wrote: > I compiled a list of countries which recently gained independence... Generating my prediction involved estimating "the probability that a [typical] potential nation becomes a UN member per year". Now, it's quite easy to calculate a rate of "number of [potential] countries that become UN members per year", but you also need a "number of *potential countries that could become UN members*". How do you arrive at an a reasonable estimate for that? At first, it seemed impossible to come up with anything other than a within-a-few-orders-o...

@Linch I have very marginally updated down from my base rate because I believe COVID-19 has slightly impaired the average competitor's preparation for the event. e.g. I think many countries have their teams coached by former competitors and other mathematicians, and I expect that there's been fewer coaching sessions than usual this year. Although maybe being in lockdown has given people more time to prepare...

@(tenthkrige) Here's my model: I think many new or less committed Metaculites look at a lot of the questions presented to them and go "Jesus, I have no clue how to predict that", skip those, and instead pick questions that seem "easier" to predict on. (cf. the Trump 2020 question.) Now, this question has more predictions than average, and I suspect it's because it's seen as one of the "easier" ones by such users. Even a relatively uninformed person has seen a bunch of headlines over the years saying "2014 was the hottest year on record", "2015 was the h...
@(Linch) So I'd heard this as a factoid in various EA/rationality media ("[when the authors rounded the superforecaster’s forecasts to the nearest 0.05, their accuracy dropped](https://forum.effectivealtruism.org/posts/W94KjunX3hXAtZvXJ/evidence-on-good-forecasting-practices-from-the-good)") and I think Tetlock's interview on the 80,000 Hours podcast, but actually chasing up this claim down to the [primary source](https://academic.oup.com/isq/article/62/2/410/4944059#118506527), it says "[The average superforecaster could reliably parse her judgments int...

Rewards as an incentive for authoring questions?

As a general point: More questions in the system is surely a good thing. Only < 300 binary questions have been resolved in Metaculus's lifetime. I feel that to get a real sense of the calibration of the community, of the best individual predictors, and of Metaculus's prediction formula, the sample size of resolved questions should ideally be 10x or 100x larger than it currently is.

[There have been 48 Vice Presidents including Mike Pence.](https://en.wikipedia.org/wiki/Vice_President_of_the_United_States) ["Of the 14 vice presidents who fulfilled their ambition by achieving the presidency, eight succeeded to the office on the death of a president, and **four of these were later elected President**."](https://www.senate.gov/artandhistory/history/common/briefing/Vice_President.htm) Ford also became President after Nixon's *resignation* rather than his death, but did not win the following election. This leaves 5 who were elected to t...
China officially ["confirms birth of gene-edited babies"](https://www.scmp.com/news/china/science/article/2182964/china-confirms-gene-edited-babies-blames-scientist-he-jiankui) in a [short bulletin](http://www.xinhuanet.com/local/2019-01/21/c_1124020517.htm) by the state-run Xinhua News Agency. Antonio Regalado, [who first broke the story](https://www.technologyreview.com/s/612458/exclusive-chinese-scientists-are-creating-crispr-babies/) (prompting the Associated Press and He Jiankui himself to reveal the experiment's existence earlier than they both pl...

@ignorance.prior There's good evidence that the kind of antibody test used in Miami-Dade has a false positive rate of 13% (not a typo). See Twitter thread here, and associated COVID-19 Testing Project website here.

From the community distribution, it looks like a lot of people are basing their predictions on a linear forecast. There are two observations that make me think it's pretty likely that a linear forecast is going to overestimate the number of cases: * In the simple linear regression that I performed using data points from April 1st to 27th, the residuals vs fitted values curve shows that the model underestimates in the middle of the range and overestimates at the ends, indicating a curve that's flattening * The number of new cases today is lower than any...

I would like to complain about the lower bound for the slider. It's far too high. I'm not able to make my honest prediction, which is that the entire population of the Earth will enter space by 2 PM tomorrow.