"EA residencies" as an outreach activity

by Buck 25d17th Nov 20196 min read8 comments

83


[This was partially inspired by some ideas of Claire Zabel's. Thanks to Jessica McCurdy, Neel Nanda, Kuhan Jeyapragasan, Rebecca Baron, Joshua Monrad, Claire Zabel, and the people who came on my Slate Star Codex roadtrip for helpful comments.]

A few months ago, some EAs and I went on a trip to the East Coast to go to a bunch of Slate Star Codex meetups. I'm going to quote that entire post here (with a couple edits):

Our goals are:
- to meet promising people at the SSC meetups and move them into the EA recruiting pipeline
- to spend some time with promising new EAs, eg those at student groups, in the hope that spending a few hours of focused one-on-one time with one of us will help get them more into EA. Like, I think 80K finds people who are excited about AI safety stuff but aren't very knowledgeable about it yet; I think that those people can maybe get a lot out of a few hours' conversation with a few people who have worked professionally on this stuff.
- to visit EAs who are "in holding" doing things like PhD or EtG tech jobs, with possible good outcomes being that they'll be fired up wrt EA and more likely to do really impactful EA stuff on a timescale of like a year, or that their improved (Bay Area/professional EA) connections make it easier for them to spot good opportunities or move into doing more impactful work.
- (less primary) to talk to hardcore EAs and swap arguments and get to know each other better
Here's why I think it's worth us talking to various promising new EAs and enthusiastic EAs who haven't worked in the EA scene full time:
- There are a lot of accumulated arguments about EA topics which I think it’s really helpful to think about but which are hard to access when you only know EAs on the internet, because those arguments haven't been written up clearly or at all, or because their writeups are hard to find and rely on background knowledge that you don't know how to acquire.
- A lot of the time, EAs present versions of arguments that are strong enough to convince you to tentatively think that it's worth engaging seriously with the possibility that the conclusion might be true, but which have a bunch of holes in them that require substantial thinking to fill. Sometimes EAs (eg me) make the mistake of conflating these two levels of strength of argument, and act as if people should be persuaded by the initial sketch. One way that I notice when I'm making this mistake is by getting in arguments with people who've thought about stuff more than me. I hope that talking to more knowledgeable EAs might help some of the EAs we hang out spot holes in their understanding that might help improve their understanding and their epistemics.
Here is a reason that I think that having SF Bay Area EAs talk to rationalists in these cities at SSC meetups is plausibly worthwhile:
When smart people are skeptical of some of my weird beliefs, eg that AI x-risk is really important, or that they should consider working on EA stuff, or that long term we should consider radically restructuring the world to make it better for animals, a lot of the time their disagreement stems from something true about the world that the arguments they've seen didn't address. This is hard to avoid because if you try to write an argument that addresses all the potential concerns, it will be incredibly long. But this makes me think that it's often really high impact for people who have thought a lot about these arguments to talk to people who have heard of them but felt very unpersuaded.

My predictions mostly matched my impressions of what happened.

But I think you might be able to get many of these benefits more efficiently by doing something more like a residency, where you spend a relatively long time in each city, compared to doing a tour, for a few reasons:

  • If you want to spend 20 hours talking to people in a city, you can do this more efficiently if you choose the best twenty hours in a ten day period, rather than having to do 20 contiguous hours. We missed out on a bunch of good opportunities because we were only in each city for less than 24 hours. It meant that if people weren't free right when we were in the city, we couldn't talk to them.
  • Spending more time in a city means that your travel costs are amortized over more hours.
  • If you only do a few hours of talking to people each day, you might be able to do it with your social time budget, and not sacrifice many hours of normal work.

Here's an example plan:

  • For a month, Alice does her usual EA work from the US East Coast.
  • She stays mostly with EAs who she wants to spend more time getting to know.
  • She tries to get quite a lot of normal work done while she’s there.
  • She tries to meet a bunch of EAs/EA-relevant-people in her evenings. She goes to meetups and parties. Maybe she tries to get people to host a few extra meetups and parties. (If she has some kind of reputation that allows her to draw a crowd, this can also benefit the local EA groups.)
  • She stays mostly in Boston, but she also spends a weekend in New York and spends a few days at Yale or at other places with a bunch of promising EAs. (I think it can be nice to be in a place twice, so that the first time you meet people and the second time you meet the people those people wanted you to meet, and you can see some people twice, which I think is often pretty helpful, so she does.)

Here are the main costs:

  • Logistical inconveniences
    • travel time
      • hopefully this isn't too bad if you can take trains (on which you can work), which you can do if you’re not in a hurry. One lesson from the SSC roadtrip is that travel is substantially more inconvenient when you have time pressure.
    • organizing events and conversations with people
  • Work efficiency penalties.
    • You might sleep worse. You could get around this by sleeping in hotels, or being vigilant about ensuring a good sleep environment when sleeping at the houses of EAs.
    • Some types of work are a lot more efficient with in-person collaboration. Maybe your job doesn’t involve that kind of work, or maybe you could arrange to not do that kind of work for a month.
    • You might have a worse work environment. You could get around this if you made sure you had eg a large monitor everywhere you were going to be trying to work. Planning to work out of a WeWork or something could also be helpful for this.
  • Most people don’t like being far away from home for long

Here are the properties that I think make someone a good candidate for this, other than being an EA with interesting experience or perspectives or similar:

  • It seems good if you’re generally likeable, and interesting in conversation, and have a good sense of social appropriateness in unfamiliar situations. Neel Nanda, a Cambridge student who IMO has good judgement about EA outreach, said: “I have a slight apprehension that an outsider from a more established community trying to do outreach in smaller ones could come across as patronising/arrogant. I'd also be very concerned about the person coming across as weird (essentially the unfamiliar situations point)”. I think this is a crucial point.
    • One obvious subtlety here is that different people vibe differently well with each other. Neel says, “I'd also expect cultural differences to be quite important and a potential source of failure modes, especially if someone has well calibrated social skills in their social context but doesn't properly account for the change in context, eg a Bay Area EA coming to the UK would probably come across as very outgoing, potentially offputtingly weird/arrogant/confident, and I think the baseline level of social confidence eg requesting 1 on 1s probably differs a lot”.
      • You could try to address this concern by going with someone with complementary social skills
    • I think it’s also much easier to pull this off if you understand things about the local EA scene like their interests and culture, and if you know some local members reasonably well already, so that you can get rapid feedback from them about how this is going.
    • I think that this probably works much better if the EA in residency isn’t trying to represent all of EA, they’re just trying to represent themselves, as an EA who has opinions about things, and they make it clear that they are not a representative of all of EA. If you do this, you’re less making a claim about your own legitimacy, you make it clearer that you’re not speaking for all of EA (which frees you up to share your nonstandard EA opinions), and people might jump less to the conclusion that all EAs have the same beliefs as you.
  • It’s good if talking to strangers is more often fun and fulfilling than stressful or tiring. You could try to set it up so that you were mostly talking to people in ways that aren’t as tiring, by for example mostly talking to people who you feel you can relax about.
  • Having work that you can do fairly effectively remotely.
  • Having a lot of familiarity with the EA community and EA cause areas, and a lot of enthusiasm for talking and thinking about this stuff.
  • I think that being interested in EA outreach is somewhat helpful, because a lot of the people you talk the most to do EA outreach work (eg running EA groups) and are interested in talking about it.
  • Being good at quickly getting a sense of people, so that you can eg spot whether this person should be introduced to a particular person you know (being well-connected also helps for this).

I think that it might be good for people to do residencies like this. If you're interested in doing this, I'm interested in talking to you about it. I'm also interested in talking to people who live in places which they think would benefit from this type of thing.

83