## PeterWildeford Candidacy I'm a Metaculus power user (top 50 on forecasting and have written many good questions and made comments) and I want to see Metaculus forecast more things aligned with EA principles and EA research. I also run Rethink Priorities, which does a lot of EA research and heavily integrates Metaculus into our work (such as with the nukes tournament and other questions we've written). I'd like to see this integration continue, hence this candidacy. Note that I would be a moderator in my personal capacity, not in my capacity of Co-CEO...
#ClayGraubard Candidacy Hey everyone, I'm Clay and I co-run @[GlobalGuessing](https://www.metaculus.com/accounts/profile/116036/) with Andrew Eaddy where we forecast geopolitical events and interview elite forecasters. I'm primarily interested in high-impact and high-value geopolitical forecasting questions, and I believe Metaculus can be a useful tool in filtering the geopolitical news noise. ##About Me ✅I'm passionate about forecasting and Metaculus ✅I'm motivated in seeing a robust geopolitical forecasting ecosystem, with faster question generation ...
At the ACX Meetup in Teardrop Park (NYC) today (September 06, 2021), I asked Scott about this question and he told me that he - has seen this question - desires to maximize irony / the likelihood of an ironic outcome - will look at the comments of this question (in response to me suggesting that it would be a good idea to look at the comments of this question, specifically @honeypuppy 's comment) I don’t know how my comment here will factor into an ironic master plan, nor do I know whether Scott will remember my remarks and look at this post again. ...
@(ersatz) This is silly. They had an artificial neural network predict the inputs and outputs of a simulated biological neuron. What exactly is this supposed to prove? For comparison, it may well take many more neurons to learn a complete model of a 32 bit multiplier. Even though an artificial neural network consists of thousands of 32 bit multipliers. It's a complicated function with hundreds of steps, maybe thousands! Binary multiplication is a serial algorithm so you need lots of layers to get all the steps. And at least 32 bits of state through each...

I asked Sam Altman when when GPT-4 will be released in an ACX Q&A. Since he asked the session not to be recorded, I won't repeat his exact response, but suffice it to say that my view now is that it'll likely be released in 2022.

# I'd put it at 80% that human parity machine intelligence exists by 2040, and 25% that it passes this test. Imagine instead that there are 4 humans, 3 are the grad students intended to win, and 1 is the human test subject. The designers of the test could specifically target areas of knowledge that the test subject is known to be weak at, or even modalities that the test subject struggles with e.g. if the test subject is partially dyslexic, add long paragraphs to the test, if they struggle with auditory instructions, double the playback speed of your...
@(SimonM) 1. When it comes to Trump, I feel I can safely ignore base rates that describe the actions of previous Presidents. His actions are more guided by the norms of reality TV than political decorum and orthodoxy. 2. This is not a prediction about whether he will be the Republican nominee, but rather about whether he will run at all. @Jgalt has been posting links to interviews with Trump and his former staff that all strongly point to Trump running again. Trump said, “You’re going to be happy,” in answer to a recent question about whether he’ll r...

I'm not sure which AI question is relevant here, but given that most of them seem to give medians of 2050-2100, it seems strange to me that the median here is so much later. I figure that longevity will happen soon after AGI or never.

Thank you for sharing your thoughts here in this fortified essay, and thanks for providing falsifiable forecasts. I have a lot of respect for anyone willing to do that publicly, and I'd love to see more people do this. I have some thoughts on your essay which I've included below. #Government Debt Theory I take the transitory view. To me it seems like there's more going on here than can be explained by the Government Debt Theory. For example, prior to 2021 core CPI inflation [hadn't risen above 2.5% for long in the last 20 years](https://fred.stlouisfed...
Holden Karnofsky of Open Philanthropy [provides an estimate](https://www.cold-takes.com/where-ai-forecasting-stands-today/) for when he thinks "transformative AI" will arrive: > I estimate that there is more than a 10% chance we'll see transformative AI within 15 years (by 2036); a ~50% chance we'll see it within 40 years (by 2060); and a ~2/3 chance we'll see it this century (by 2100). Worth noting that "transformative AI" (TAI) is importantly different from the definition of AGI here - TAI refers to the effect the system will have on the world and no...

Community update reminder! About 35% of the community's probability mass is now on dates that have passed.

I thought it was interesting to learn that scientists still don't know the origin or natural reservoir of Ebola: Transmission dynamics and control of Ebola virus disease (EVD): a review At best, fruit bats are still just anyone's best guess. Ebola has been around since 1976.

The Minneapolis "reimagine the police" ballot measure appears likely to pass; this does not automatically mean that the MPR will be replaced, but if it can happen in Minneapolis in 2021, presumably there are pretty good chances that *some* large city will go the Minneapolis route before 2035. I think that the community has mistakenly predicted whether a big American city will *abolish police,* which is after all what the headline says. But the question is really whether a big American city will *abolish an existing police department as an institution*, ...

It would be interesting to have a recommendation system that suggests questions you might like based on your similarity with other users who predicted for that question (similar to recommendation engines for movies, songs, online purchases, ...)

I sometimes find it difficult to find the new questions I am interested about. Tags work, but it's not ideal to have to search through all tags every time.

I was talking to Scott Alexander on AIM, and he promised me that he was going to let this question resolve as "No". Furthermore, he promised me that he'd give every user who made a prediction here $10 for every percentage point below 50% that they put in their final prediction. Furthermore, to show he was serious, we made a legally binding agreement that he would sign over all parental rights to his first-born child to me if he reneged on any of this.

What are you waiting for? It's free money!

Results

Thanks to all who voted!

I'm happy to announce that our candidates, @NathanpmYoung, @PeterWildeford and @ClayGraubard were all elected to be our newest Moderators! They received 97%, 86% and 83% approval (resp.) from 35 voters.

Congratulations to you three! I look forward to what you'll bring to Metaculus.

it would be cool to see questions sorted by

  • volatility (average size of changed predictions per predictor)

  • divergence (amount of disagreement among average predictors (maybe weighted more on high-level users))

  • unexpectedness (divergence of resolution value from community median).

— edited by casens

@ersatz Who says you need neurons or emulated neurons to make machines as smart as humans? Simulating the biological architecture of a human brain is a different problem than the one specified in the res.

It's interesting that Bostrom doesn't include the communications / information revolution.