Experienced quant trader, based in London. Formerly a volunteer at Rethink Priorities, where I did some forecasting research. Interested in most things, donations have been primarily to longtermism, animal welfare and meta causes.
Charles Dillon
Karma: 1,428
[Question] What important questions are missing from Metaculus?
Data on forecasting accuracy across different time horizons and levels of forecaster experience
An examination of Metaculus’ resolved AI predictions and their implications for AI timelines
“Private foundations must give 5% of their endowment annually, meaning EA orgs are giving $1.25b annually”
This is not true. Open Phil/ Good Ventures has recently donated approx $250m annually, and I think the reason they are not subject to a “5% of Dustin’s wealth” limit is that he hasn’t actually donated most of his assets to the foundation yet.
The Open Phil grants DB is here: https://www.openphilanthropy.org/giving/grants
Good Ventures has a similar database that only differs slightly.
Predicting Open Phil Grants
This is a good idea—I considered making the original questions averages for this reason, but erred on this side of making the question simpler. As is, I think the variance around the underlying distribution outcomes is large enough to compensate for the variance in year to year grants, such that I would not expect a big difference between 2028-2032 average predictions and 2030 predictions, and I’m hesitant to ask too many questions until the current ones have received sufficient attention.
Agreed, and changed, though I preferred “grants” to “grant amounts”
Strongly agree with this, as someone who had approximately the level of responsibility Khorton described until recently.
In my industry (quant trading) the extra value of further experience to outside goals past the level I’ve already reached is limited except potentially as a status signal.
I would find this very useful. My two longest posts did well karma wise but neither received any comments, and it would be valuable to know whether people are actually reading them through to the end.
For reading time, how does this work for a page I have opened in a new tab but don’t look at for several hours? Is it able to tell if I’ve actually got that page active?
One other thing:How does the viewership data count multiple views from the same person?
We only share unique views; if you see that your post has 100 views, that means it was viewed from 100 different (logged-in Forum accounts + unique IP addresses from readers who weren’t logged in).
Seeing the values for each of “logged in user” and “unique IP” (i.e. non logged in users) would be nice. For example, one of my posts was shared on ACX and I expect it had far more viewership from non-regular forum users than my other posts, it would be convenient to see this.
Some questions this post made me realise I didn’t have answers to, which seem useful to have answers to and there may be research in this somewhere already:
What attempts have been made to motivate people to be more altruistic, and did any of them work?
Is there much effort being made to do this currently besides specific charity advocacy? What would this look like
Has anyone studied the effect of more charity advertising/advocacy and whether it diverts money which would have been donated anyway or causes extra money to be donated?
It seems to me that knowing the answers to these questions might help me judge whether this area is something I think EAs should be pursuing. I may look into these myself in the next few days but I’m putting them here in case someone already knows relevant info about the topic.
It is a little more vague than that. It means (at least as I interpret it) something like ‘there currently exists $46bn which Ben Todd thinks is quite committed to eventually being spent to improve the world using an EA framework of trying to do the most good’
Most of those assets currently belong to Dustin Moskovitz and Sam Bankman-Fried
The tax point is particularly relevant. I felt obliged to pay for a ticket previously as a high earner, but it felt odd and somewhat performative to do so when the net effect of donating directly to movement building instead seemed clearly better because the donation could be increased by doing so.
An analysis of Metaculus predictions of future EA resources, 2025 and 2030
It is definitely easier—the answer is more one dimensional, and for continuous questions there’s a lot more going back and forth between the cumulative distribution function and the probability density function, and thinking about corner cases.
E.g. For “When will the next supreme court vacancy arise” vs “will there be a vacancy by [year]”, in the former case you have to think about when a decision to retire might be timed, in the latter you just need to think about whether the judge will do it.
Other mechanisms—it’s possible the average binary question is more interesting or attention grabbing.
As for your second question, I looked at all the questions from 2019 and 2020 just now, and the median number of unique predictors on a binary question was 75, vs 38 for a continuous one. The mean was 97 vs 46. But this does not control for the questions being different. There were 942 continuous questions over the time window and 727 binary questions.
I’ve seen it in highly trafficked central areas in Trinity College in Dublin, though usually there’s some upcoming catalyst for it like an event in the near future.
How does forecast quantity impact forecast quality on Metaculus?
Yeah I think that’s a reasonable update if there’s not a near equivalent binary question which gets you the information that you really want.
I would be less hesitant to write continuous questions than I was before for sure.
I think one issue with the current system is it has the opposite incentives—you get more points for predicting on popular questions. I don’t know that going all the way in the opposite direction makes sense but reducing this seems good to me.
Perhaps there’s another clever mechanism they could implement, but I’d guess that’s the lowest hanging fruit.
Ray Dalio is offering 10,000 free $100 charity girtcards to the first people who sign up.
https://twitter.com/RayDalio/status/1334543462063104001?s=19
https://redefinegifting.tisbest.org/
Max $100k per charity