electroswing

Karma: 520

electroswing 27 Mar 2023 19:48 UTC
4 points
1 ∶ 0
on: electroswing’s Shortform
On the EA forum redesign: new EAs versus seasoned EAs
In the recent Design changes announcement, many commenters reacted negatively to the design changes.
One comment from somebody on the forum team said in response: (bolded emphasis mine)
One of our goals on the Forum team is to make the Forum accessible to people who are getting more engaged with the ideas of EA, but haven’t yet been part of the community for a long time.. Without getting into a full theory of change here, I think we’ve neglected designing for this user group a bit over the last several years. Some of the barriers to entry for these folks include:
- Feeling that the Forum experience (fonts, look and feel) is quite jarring, and different from a lot of the internet they’re used to.
- Understanding what the Forum as a space is all about
This feels like a crux. Personally I think the EA forum should be a place seasoned EAs can go to to get the latest news and ideas in EA. Therefore, making the EA forum more similar to “the internet [new EAs are] used to” should not really be a priority.
There are so many other spaces for new EAs to get up to speed. It’s not obvious to me that the forum’s comparative advantage is in being a space which is especially welcoming to new users.
To my knowledge, this tradeoff between designing UX for new versus seasoned EAs has not been publicly discussed much. Which is a shame, because if the EA Forum is a worse space to exist in for seasoned EAs, then seasoned EAs will increasingly retreat to their local communities and there will be less interchange of ideas. (e.g. think about how different Bay Area EAs are from DC EAs)

electroswing 11 Jan 2023 19:27 UTC
7 points
3 ∶ 0
on: Non-trivial Fellowship: start an impactful project with €500 and expert guidance
Question about timing of program: This program is during the school year rather than during summer break. It is also meant for EU/UK students, who may not have slack during the school year because EU/UK university admissions often specifically require very high grades with little room for error. Do you think your application pool would be stronger if this were a summer program instead?
(pointed) Questions about “puzzle quiz”:
1. These synonym questions (and to a lesser extent the analogy questions) are dramatically easier for native English speakers. What about EU students who are non-native English speakers?
2. The test displays your results at the end (below is an excerpt of my results from clicking randomly). If you give students a number, that’s going to give them much more reason to Goodhart the shit out of it, speculate about cutoffs, etc, etc. Also a person might simply feel inadequate looking at the results of what is essentially an IQ test heavy on English vocab and probability theory. Do you have a reason for displaying these results?

electroswing 10 Jan 2023 23:45 UTC
1 point
0 ∶ 0
on: electroswing’s Shortform
https://www.aeaweb.org/articles?id=10.1257/aer.20141734 Gender Differences in Accepting and Receiving Requests for Tasks with Low Promotability ”
“We examine the allocation of a task that everyone prefers be completed by someone else (writing a report, serving on a committee, etc.) and find evidence that women, more than men, volunteer, are asked to volunteer, and accept requests to volunteer for such tasks.”
Promotability isn’t exactly the word that applies to EA. Instead here I mean a more nebulous term like “low promotability = grunt work, lack of prestige, lack of career capital outside of EA, lack of intellectual labor, lack of leadership displayed, lack of skills built...”
The people who do operations, event planning, and personal assistant work in EA are disproportionately women+nb. And then roles on the opposite end of the spectrum like “Independent AI Safety Researcher” are disproportionately men. Anecdotally, I see university-aged women+nb taking time away from their studies to do community building, and university-aged men taking time away from their studies to do upskilling or research.
There’s nothing wrong with ops, event planning, and personal assistant work, but I worry highly qualified women+nb are selling themselves short.

electroswing 27 Nov 2022 1:11 UTC
3 points
1 ∶ 0
on: What videos should Rational Animations make?
Short productivity videos. For example: “what is a TAP”, “what is Murphyjitsu”, “what is goal factoring”, etc.
This, alongside your current “worldview expanding” content, curates an audience who is interested in tackling big questions and also cares about optimizing their personal impact.
Expanding on the “big questions” side, I would like to see more content which inspires altruism (example).

electroswing 10 Nov 2022 16:03 UTC
2 points
0 ∶ 0
on: I there a demo of “You can’t fetch the coffee if you’re dead”?
There is a relevant Rob Miles Computerphile video. It does not have a demo component like you are planning, but it does seem to click with laypeople (1M views, top comments generally engaged).

electroswing 8 Nov 2022 19:53 UTC
1 point
0 ∶ 0
on: What is the right number of EA-aligned donation advisors?
Two more organizations which try to get a narrow group of people (athletes / founders) to give to effective causes:
- High Impact Athletes
- Founders Pledge
One more organization trying to get the everyday person to give 1%:
- One for the World

electroswing 8 Nov 2022 15:56 UTC
6 points
0 ∶ 0
in reply to: henryj’s comment on: EA Groups Should Rename Intro Fellowships for Better Clarity
Perhaps one source of downvotes is that the main idea of this post is unoriginal. Anyone putting on an intro fellowship has put some amount of thought into:
- Do I call it a “fellowship” to give it prestige, or do I call it a “seminar” / “reading group” to make it sound academic, or do I call it a “program” or a more neutral tone, …
- Do I call it “Arete” to sound fancy, or do I call it “intro” to sound welcoming, …
- Do I explicitly put “EA” in the title?
The one new thought here seems to be having the acronym “IDEA” stand for “Introduction to making a Difference through Effective Altruism”. And this post isn’t even a comprehensive exploration of the pros and cons of this acronym! OP leaves off one important downside: at other universities (e.g. Harvard, Brown), IDEA stands for “In-Depth EA”. Another point OP doesn’t expand on is why “fellowship” had religious connotations for so many people. (Could this be more of an issue in LSU / the US South in general, compared to other parts of US/Europe?)
Finding good names is important, but this post doesn’t really do much in aim of this goal. The pros and cons here are exclusively supported by anecdotal evidence or OP’s personal aesthetics. Stuff I would like to see in the “naming CB interventions better” space:
- More experiments such as “Should you use EA in your group name?” An update on PISE’s naming experiment
- A compendium of names used at different universities / regions, with attached retrospectives from organizers
- (Probably too high effort) Proper user testing on group names

electroswing 7 Nov 2022 16:30 UTC
12 points
1 ∶ 0
on: EA orgs should accept Summer 2023 interns by January
See also: https://forum.effectivealtruism.org/posts/bmwwQoFznjRwswCbz/announce-summer-ea-internships-farther-in-advance

electroswing 29 Oct 2022 0:05 UTC
5 points
0 ∶ 0
on: What should I ask Ajeya Cotra — senior researcher at Open Philanthropy, and expert on AI timelines and safety challenges?
What does she think policymakers should be trying to do to prevent risks from misaligned AI?

electroswing 25 Oct 2022 13:45 UTC
4 points
1 ∶ 0
on: electroswing’s Shortform
Now that Rational Animations has the human capital, budget, and experience to make high quality videos like this one, I think they should develop a more consistent brand.
They should have a consistent single face or voice of the channel. Popular edutainment channels often take off when viewer connects with a likeable personality. Examples:
- Tom Scott, VSauce, Veritasium, Physics Girl, …
- Channels which don’t show their face in their typical format: Wendover Productions, 3Blue1Brown
- Even high-budget channels like Vox are starting to lean into this format by structuring their videos more like vlogs, where the viewer connects with the presenter. example
Also just look at the comments of these videos. People engage with the content, but they also feel connected to the person presenting, and write things like “Wow I liked how excited <presenter> got when <thing> happened”.
They should mark as private or remake the old videos without Rob Miles as narrator. Personally, the old videos are a bit jarring to click on—sometimes you get a guy with an accent and a bad mic (one is ok, two makes a video difficult to understand), sometimes you get a generic overly cheery American “radio voice”.
Maybe get rid of the dogs/cats? Looking at the last year of videos (there are 8), the top 5 most viewed do not have dogs/cats in their thumbnail, and the bottom 3 do. YouTube allows for extensive thumbnail A/B testing and so if they’re not doing this already, Rational Animations really should prepare more kinds of thumbnails and optimize for getting people to click on their videos (in a truthful way). Personally, when I first visited the channel, I found the dogs/cats in the Bayes video off-putting (“why are dogs/cats here? did a 12 year old girl draw this?”), but I thought they were fine in the How to Take Over the Universe (in Three Easy Steps) because they were subtle and the animation felt cohesive overall.

electroswing 11 Oct 2022 4:17 UTC
4 points
1 ∶ 0
on: EAforum badly needs its own meme page
What about a subreddit?

electroswing 9 Aug 2022 18:46 UTC
3 points
0 ∶ 0
in reply to: Bella’s comment on: Nice 1min 80,000 Hours ad on YouTube video
If it’s OK to answer: how much did you end up dictating the content/wording of the ad?

electroswing 23 Jul 2022 18:54 UTC
7 points
0 ∶ 0
on: The EA forum post-writing algorithm (95% > 100 karma, n> 100)
This might be better received as an April Fools’ Day post.

electroswing 16 Jul 2022 13:07 UTC
1 point
0 ∶ 0
in reply to: juliakarbing’s comment on: Think about EA alignment like skill mastery, not cult indoctrination
I should clarify—I think EAs engaging in this behavior are exhibiting cult indoctrination behavior unintentionally, not intentionally.
One specific example would be in my comment here.
I also notice that when more experienced EAs tend to talk to new EAs about x-risk from misaligned AI, they tend to present an overly narrow perspective. Sentences like “Some superintelligent AGI is going to grab all the power and then we can do nothing to stop it” are thrown around casually without stopping to examine the underlying assumptions. Then newer EAs repeat these cached phrases without having carefully formed an inside view, and the movement has worse overall epistemics.
Here is a recent example of an EA group having a closed off social circle to the point where a person who actively embraces EA has difficulty fitting in.
Haven’t read the whole post yet but the start of Zvi’s post here lists 21 EA principles which are not commonly questioned.
I am not going to name the specific communities where I’ve observed culty behavior because this account is pseudoanonymous.

electroswing 7 May 2022 1:56 UTC
9 points
0 ∶ 0
on: SERI ML Alignment Theory Scholars Program 2022
I worry that the current format of this program might filter out promising candidates who are risk averse. Specifically, the fact that candidates are only granted the actual research opportunity “Assuming all goes well” is a lot of risk to take on. For driven undergraduates, the cost of a summer opportunity falling through is costly, and they might not apply just because of this uncertainty.
Currently your structure is like PhD programs which admit students to a specific lab (who may be dropped from that lab if they’re not a good fit, and in that case, will have to scramble to find an alternative placement).
Maybe a better model for this program is PhD programs who admit a strong cohort of students. Instead of one two-week research sprint, maybe you have 2-3 shorter research sprints (“rotations”). From a student perspective this would probably lower the probability of them being dropped (since all of the mentors would have to dislike them for this to happen).
What you’re currently doing seems like a fine option for you with little downside for the students if:
1) “Assuming all goes well” means >90% of students continue on with research
2) The projects are sufficiently disjoint that it’s unlikely a student is going to be a good fit for more than one project (I think this is probably false but you know more than me, and maybe you think it’s true)
3) 2-week research sprints are much more valuable than 1-week research sprints (I am not convinced of this but maybe you are)
If not all of these are the case I argue it might be better to do rotations / find other ways to make this less risky for candidates.
Other idea to avoid filtering out risk averse candidates: You could promise that if they don’t get matched with a mentor, they can at least do <some other project> , for example, they could be paid to distill AI Safety materials.

electroswing 6 May 2022 22:51 UTC
5 points
0 ∶ 0
on: Introducing the ML Safety Scholars Program
Can undergraduates who already know ML skip weeks 1-2? Can undergraduates who already know DL skip weeks 3-5?

electroswing 4 May 2022 21:49 UTC
8 points
0 ∶ 0
on: Introducing the ML Safety Scholars Program
You may already have this in mind but—if you are re-running this program in summer 2023, I think it would be a good idea to announce this further in advance.

electroswing 4 May 2022 5:27 UTC
7 points
0 ∶ 0
on: Most problems don’t differ dramatically in tractability (under certain assumptions)
I was in the process of writing a comment trying to debunk this. My counterexample didn’t work so now I’m convinced this is a pretty good post. This is a nice way of thinking about ITN quantitatively.
The counterexample I was trying to make might still be interesting for some people to read as an illustration of this phenomenon. Here it is:
Scale “all humans” trying to solve “all problems” down to “a single high school student” trying to solve “math problems”. Then tractability (measured as % of problem solved / % increase in resources) for this person to solve different math problems is as follows:
- A very large arithmetic question like “find 123456789123456789^2 by hand” requires ~10 hours to solve
- A median international math olympiad question probably requires ~100 hours of studying to solve
- A median research question requires an undergraduate degree (~2000 hours) and then specialized studying (~1000 hours) to solve
- A really tough research question takes a decade of work (~20,000 hours) to solve
- A way ahead of its time research question (maybe, think developing ML theory results before there were even computers) I could see taking 100,000+ hours of work
Here tractability varies by 4 orders of magnitude (10-100,000 hours) if you include all kinds of math problems. If you exclude very easy or very hard things (as Thomas was describing) you end up with 2 orders of magnitude (~1000-100,000 hours).

electroswing 28 Apr 2022 22:39 UTC
9 points
0 ∶ 0
on: ‘Dropping out’ isn’t a Plan
I think the diagram which differentiates “Stay in school” versus “Drop out” before further splitting actually has some sense. The way I read that split is, it is saying “Stay in school” versus “Do something strange”.
In some cases it might be helpful, in abstract, to figure out the pros and cons of staying in school, before recursing down the “Drop out” path. Otherwise, you could imagine a pro/con list for ORGs 1-3 having a lot of repetition: “Not wasting time taking useless required classes” is a pro for all 3, “Losing out on connections / credential” is a con for all 3, etc.

electroswing 30 Mar 2022 15:25 UTC
3 points
0 ∶ 0
on: What’s the best machine learning newsletter? How do you keep up to date?
Yannic Kilcher’s youtube channel profiles fairly recent papers / “ML news” events. The videos on papers are 30-60mins, so more in depth than reading an abstract, and less time consuming than reading the paper yourself. The “ML news” videos are less technical but still a good way to keep up to date on what DeepMind, Meta, NVIDIA, etc. are up to.