I’m a generalist and open sourcerer that does a bit of everything, but perhaps nothing particularly well. I’m currently the AI Safety Group Support Lead at CEA.
I was previously a Software Engineer in the Worldview Investigations Team at Rethink Priorities.
This is of course, pretty much the conclusion of Yudkowsky’s classic Purchase fuzzies and utilons separately.
But the main lesson is that all three of these things—warm fuzzies, status, and expected utilons—can be bought far more efficiently when you buy separately, optimizing for only one thing at a time
The mentioned tweet seems to no longer be available. Could you provide screenshots?
Does anyone have any idea of how politically feasible it is for the bill to pass?
I’m thinking that many of these issues with the forum are fixable directly within the forum:
Newbie-friendliness: This is probably the hardest to fix directly, but this seems more related with the EA culture than anything platform-specific.
SEO: Nothing about the forum seems to put it at a disadvantage in terms of SEO, and in my experience, Q&A posts already show up in search results.
Data and Statistics: I’m not aware of how the analytics system of the EA forum works, but in general, there’s no clear limitation on performing makeshift analyses on current forum threads, see this, by example.
Overall, I think it’s difficult to justify a new platform, especially considering that splintering the community conveys pretty significant disadvantages. Specifically, creating a new platforms risks creating a new information silo and reducing newbie involvement with the rest of the community (while not allowing a clear path for them to start engaging with the forum).
Are we including 80k’s problem profile on Climate Change here? This is the explanation that is included in the handbook (and in the intro fellowship) seemingly, precisely for this reason.
While I agree in general, the problem is “something shorter, with more direct comparisons to other cause areas” might have the opposite effect. That is the kind of argument that could induce emotional rejection on people that have already spent significant resources (or have modeled their identities) on fighting climate change. For that specific group of people, you probably need something with significantly more nuance.
Partly unrelated: at first, I thought the title meant that we should research deprioritizing external communication. It took me a while to understand it meant that research is/was deprioritizing external communication.
Thank you! This gave me an unexpectedly detailed peek into the Diversity Visa Program, and I hadn’t found good information elsewhere.
I love this, it would work like the Toucan extension, which replaces random words to another language, with frequency depending on difficulty.
Just wanted to point out that the 80,000 hours problem profile on Climate Change (published on May this year) might fit what you’re describing. I still think there are significant improvements to be made (discussed on my comment on the answers section).
I’m guessing the actual deadline is January 31st, 2023? (It says 2022)
Just wanted to point out your art is beautiful!
Looks like we just had a comment race condition. I’ll delete mine 😅
This is genuinely a public service. Thanks.
Related question to the one posed by Yadav:
Does the fact that OpenAI and DeepMind have AI Safety teams factor significantly into AI x-risk estimates?
My independent impression is that it’s very positive, but I haven’t seen this factor being taken explicitly into account in risk estimates.
To what degree is having good software engineering experience helpful for AI Safety research?
For what it’s worth, my independent impression was that the fact that “fellowship” sounded more prestigious was probably a good thing, mostly because it could fit within an academic culture very well[1].
I had never made the connection to religious fellowships[2], and I don’t quite understand the argument from epistemics[3]. However, I do like the potential emphasis on doing and learning, which, I think, is a problem in the current form of most introductory fellowships (see here).
This might be very dependent on the communications strategy of each group. At our local student group, we found the connection with academia helpful; however that also raises legitimate concerns with diversity.
This might be cultural; my first language is spanish, where “fellowship” is mostly used untranslated and only in the academic sense.
Why would “fellowship” point to a set of dogmas but “seminar” wouldn’t?
Is there a plan to run this program virtually anytime soon? I would love to see this as part of EA Virtual Programs.
We shouldn’t try to shut down a nonprofit based on kin associations. This would basically amount to kin punishment, and it would create a dangerous ethical precedent.
As for funding, probably most nonprofits in the community have received FTX funding directly or indirectly as of right now, and I don’t see how that should affect how we evaluate their work.
When I got involved in kick-starting our local student chapter, I noticed most of our ideas initially drifted to some form of influencing, but we ended up “correcting” that to what has become an internal motto: quality over quantity. While I still think it’s a good initial strategy for a student chapter, your argument did make me think about missed opportunities in influence.
For example, I was recently offered the opportunity to help build the syllabus for an Ethics in Computer Science course, as well as helping create social responsibility modules for an Intro to Econ course. My initial reaction was to prioritize the student chapter, but I can now see a potential opportunity to align both.
I think you’re right on the premise that there’s a way to influence people that simultaneously doesn’t run the risk of value drift or unintentionally misrepresenting the EA community to the world; this is probably more in line with traditional education, campaigning, lobbying, and activism. In my (limited) experience in the community, there seem to be many low-hanging fruits in this regard, but there have been advances in this direction, as yesterday’s post on the Social Change Lab seems to show.