I attended my first EAG in Boston last weekend and had a great time. I figured it would be appropriate to combine my first EAG with my first substantial forum post, so here goes.

Epistemic status: Everything in this post is 100% true if you are me, but results may vary if you aren't me. Everyone has personal context^[1] that influences their experience of events like this! There is also a robust body of literature on EAG,^[2] and this is only one post; weigh accordingly. Finally, although I am employed by GiveWell, this post is personal^[3] and doesn't represent GiveWell's opinion.^[4]

I. Before the conference

Conference preparation is really easy. The EAG team uses an app called Swapcard to display the agenda, push out announcements, and organize 1:1s. People love to talk trash about Swapcard, but it's almost certainly better than any other option, and all of the EA software engineers are transitioning to AI safety research^[5] instead of designing a better app, so we might as well appreciate Swapcard for what it is. Swapcard has several nice features:

1. You can look at the entire list of attendees and see information about their organizations, contact info, and topics they can help and be helped with. This is an indescribably nice feature that makes it very easy to find all the people with whom spending time would be maximally fun and useful! CEA also sends out an Excel spreadsheet with similar information, but (imo) Swapcard is better because people frequently update their profiles as the date of the conference approaches.
2. Swapcard lets you easily book 1:1s with people and set your own availability for the same. This is a very low friction feature and is extremely useful!
3. You will be so angry at Swapcard's many bugs that nothing else could possibly bother you for the remainder of the weekend. For example, I have an unread chat notification in my Swapcard. This notification is a lie (I have read all my chats), but it persists no matter how many times I refresh the app. As another example, sometimes a person will send you a meeting invite, and all the pieces of text in the invite will be superimposed on top of each other such that the invite is entirely unreadable. These and Swapcard's other bugs will make you boil with such a pure rage that no other minor irritations for the rest of the conference will rise to conscious awareness, and you will be happy.
4. Because everyone knows that Swapcard is buggy, you can use it as an excuse for basically any kind of antisocial behavior. For example, if you arrive 23 minutes late to a 25-minute meeting and say, "I'm so sorry to be late, my Swapcard was acting weird and...", the other person will cut you off and say "I know!! The same thing has been happening to me all weekend!" I haven't confirmed this, but I think there's an 80% chance that you could get away with eating non-vegan food with no social cost if you blamed the meat on Swapcard.

So, to prepare for the conference, I scrolled through the entire attendee list on Thursday, sent 1:1 requests to heaps of people, and signed up for all the sessions that seemed cool (there were many such). Easy.

II. Friday

I applied for EAG many moons ago, as I was anticipating starting a new job on GiveWell's operations team. I was very stressed about joining a competent team as a total newbie to operations, and I happened upon a forum post that said something like "Volunteering at EAG can give you great insight into solid ops work." Insight sounded pretty good; I wanted insight. So I signed up to volunteer, and my first task was to arrive at the venue at 10am Friday for two hours of training. Training was run so efficiently that we were done in 62.5% of the allotted time.

Based on the info I read on the Forum and in CEA's materials, I was expecting to work brutal 8-hour shifts under the direction of a mystic operations wizard who would push me to my limits and unlock capabilities of which I hitherto lacked the imagination to even dream. Instead, I was placed on the speaker liaison team, where I was tasked with walking speakers to their sessions, making sure their needs were met, and taking them to office hours afterward. My team leads were very nice people—they were super communicative and easy to work with on shifts, which were never more than 1.5 hours. In exchange for my volunteer efforts, I received a special blue shirt and had unlimited access to the volunteer room, which was a real gem: It was like a combined quiet working room/private luggage storage/charging station that included beanbag chairs, free food for every lunch and dinner, and toothbrushes (wow). My volunteer badge also allowed me to enter the venue as early as I wanted every day, which was quite useful.

EAG staff mentioned that they were short on volunteers. This is evidence that volunteering at EAG is underrated, because I think many more people would volunteer if they knew what it was really like. It's possible that people decline to volunteer because they think (as I did) that volunteering will consume most of their time and mental space. I think in most cases that's unlikely to happen. The breakdown in time usage on my speaker liaison shifts was something like:

• 15% focused mental effort
• 40% being a warm body who is vaguely available for questions
• 45% doing nothing, waiting around

I think most volunteers (on teams like cloakroom, room management, registration, etc.) had similar amounts of free time on their shifts, though of course some were busier. You can do a lot with that "doing nothing" time,^[6] and it's actually quite nice to have some chill downtime built into your schedule (otherwise you might push yourself too hard). I guess my point is that if the warm glow of knowing that you helped run a great EAG isn't very tempting, volunteering might still be net-positive from a purely selfish perspective! However, that warm glow is really nice. People said many kind and appreciative things to me, and I felt a sense of ownership over the smooth running of the event. You should volunteer if you go. :)

Ok, back to the conference! I made the mistake of scheduling 1:1s between the end of volunteer training (12pm) and the beginning of registration (5pm) on Friday. Pro tip: Don't do that. Apparently lots of people don't update their real 1:1 availability until Friday evening because they aren't anxious obsessive planners who review the entire attendee list on Thursday. These people will unintentionally stand you up and then abjectly apologize several hours later, which is bad vibes, not what you want. If you want to have 1:1s on Friday afternoon, message people to get confirmation before sending a booking request.

At 5pm, registration opens. Everyone gets a few items:

• Venue map. This was pocket-sized and very nicely designed.
• Shirt. Cute design in orange and black, presumably because of EAG Boston's proximity to Halloween. I saw a person walking around with an EAGxBerlin shirt that also looked cool, so I'm fairly confident that solid graphic design is a normal thing for EAG apparel.
• Badge. Unfortunately there were only two badge color options, one for volunteers and one for everyone else. Ideally, there would be four colors: (1) volunteers, (2) people interested in transitioning to technical AI safety research, (3) people who can give advice about transitioning to technical AI safety research, (4) literally everyone else. I think this system would improve the efficiency of casual social interactions by at least sevenfold.

The career fair happened on Friday for most of the evening and I attended for recruiting purposes. The organizers unfortunately decided to teach me a hard lesson about the average EAG attendee's cause prioritization by placing my booth across from the AI organizations.^[7] Otherwise, the career fair was great! Attendees seemed to really enjoy the event — I saw many lengthy conversations happening and had some good chats myself.

After the career fair, I went to a rooftop gathering that consisted mostly of folks affiliated with Charity Entrepreneurship. I already had a high opinion of CE, and after meeting several more incubatees and a few research program graduates, my estimation increased! If this sounds like a plug for Charity Entrepreneurship, that's the correct interpretation — I'm really glad the org exists and I deeply admire its work.^[8]

III. Saturday/Sunday

Tons of good stuff happens on Saturday and Sunday, I'm talking sessions, 1:1s, snacks, casual hangouts, naps, community office hours, workshops, and more.

I can't actually speak much about the sessions or workshops, because I only attended those required by my volunteer shifts. Instead, I packed my days with back-to-back 1:1s! I had 1:1s upstairs, downstairs, outside, walking, sitting, in the mall, at a bagel shop, everywhere. There are many flavors of 1:1s and most of them are great! Here are a few common types of 1:1:

• Talking about specific problems: Oh my gosh, SO useful. I'm working on a few big projects at work right now, and I discovered that there were people at EAG with direct experience on every one of them. I learned so much!!
• Forming relationships with people who do broadly similar work: For example, I met up with a lot of people who do operations and recruiting work. It's really nice to feel understood, talk about common issues/questions/cool ideas, and generally vibe out with people who are in similar positions and who think about closely related problems.
• Helping other people: If you have content in the "What I can help people with" section of your Swapcard profile, you'll almost certainly receive meeting requests to talk about that thing! For example, a few people asked me to talk about working at GiveWell and making career transitions. Note: If you work in a popular cause area or at a well-known EA org, you'll likely receive many meeting requests for advice even if you haven't explicitly said you're ok with that. It can be kind of awkward to get many such requests (in part because you just won't have the time for all of them and will need to deny a few) but try to remember that these meetings are a great way to pass on your knowledge to newer/younger members of the EA community! If you feel overburdened by a large amount of similar meeting requests, you can just create your own unofficial office hours by suggesting that everyone meet you at a particular time/place.

It is possible for 1:1s to be low quality if one or both participants don't have a clear idea of their goals for the meeting, but this failure mode is easy to avoid with careful planning. I also found to my surprise that some well-known figures within EA were actually quite accessible — in a few cases I made 1:1 requests that I thought would be swiftly rejected, but they (mostly) weren't, and I was able to have some great conversations with people whose work I've followed for years. That was really nice.

Food (vegan) was provided for several meals. The lasagna and chocolate mousse were the tastiest food items. They were so good that I didn't notice the "please don't be greedy and rude" sign until I had my second plate in hand. Whoops!

The EAG team does a really great job of organizing off-site group dinners. I ultimately didn't have the energy to attend a 30-person meal, but I passed by one of the dinner groups on my walk back to my Airbnb, and they were laughing and smiling and having a great time. :) In addition to the organized dinners, someone (idk who, but shoutout to you) made a giant Google Doc that listed many unofficial dinners and gatherings. For example, one of my friends went to an unofficial Law & AI dinner on Saturday night. I texted him to ask how the dinner was, and he said:

It's all ai & law people and the people that are law are ai law

...which I guess is just a clearcut example of truth in advertising? The dinner was apparently great fun, so congrats to the Legal Priorities Project for hosting a really nice event. There really isn't a dull moment at EAG; you can always find something to do!

IV. Afterparties and Airports

Qualy suggested on Twitter that access to the EAG nap room more than covers the ticket price, so the rest of the conference is just a free perk. I think that's correct, but since not everyone can use the nap room, a stronger argument is that the EAG ticket price is really just a reasonable cover charge for the afterparties, and the conference is thrown in for free.

EAG might actually have more afterparties than sessions. You're probably familiar with Christmas Creep, but you might not have known that EAG Afterparty Creep also exists.^[9] I personally spoke to one person who arrived at the venue late Friday evening, completed their registration, and immediately left for an afterparty. In fact, I heard that EAs in the Bay Area started having afterparties as soon as EAG was announced last week because their timelines are too short to wait. But these early afterparties are just pre-gaming for the really juicy, fresh afterparties that happen on the final day of EAG after the venue closes. Boston's Sunday night afterparties had themes including animal welfare, Halloween, general party, and kidney donation (I'm not sure if the last one counts, but it happened in the same time slot and seemed to be competitive w/ other events). I have no doubt that there was also a party for people interested in transitioning to AI safety technical research. When I last checked on the Tuesday after EAG, afterparties were still happening (not joking).

Unfortunately, my past self had already decided that being home before work on Monday was an important priority (stupid!), so I had to go to the airport instead of having shouted conversations about mechanistic interpretability for a few hours in a dimly lit college house on Harvard Square.^[10] I ate a quick dinner with a friend, which was actually really nice. We spent most of the dinner talking about how we felt Wiped, Tired, and Totally Unable To Do Even One More Thing. It's obvious in retrospect that we were both in our feels about not attending afterparties.

At my airport gate I overheard two guys arguing about the moral propriety of profiting from AI development and assumed they were from EAG. I was correct of course, and it turns out I knew them already from previous EA gatherings in my city. Sadly for them, they were also employed and couldn't attend Sunday night afterparties in far-flung locations. We got to chatting about our EAG highlights. One of the guys said that he had spent a while hanging out with someone who he'd met at a previous EAG, and he thinks they'll become good friends. The other dude said he had a really encouraging meeting about a side project in global health that he's thinking about pursuing. I said that I'd seen a close friend get recruited for a really cool job, and that I'd made some new friends and received help with some important projects. All of us had had a great time!!

V. EAG was great

I was really happy after EAG, and this is my best guess at why: I liked just about everyone I met, and I trusted them too. I had a really strong sense that everyone at the event cared in a substantial (and in some cases personally costly) way about making the world maximally good. My conversations were intellectually satisfying and quite fluid because people often understood me well without needing a lot of background info. Overall, I experienced a very delightful collection of feelings, and I think the best name for it is "community." That is, EAG helped me feel connected to the EA community in a very personal way, and that feeling alone made the conference worthwhile. I'm grateful to the team that made EAG happen, and I think I'll be going again.^[11]

1. ^{^}

    My personal context: This was my first EAG. I was a volunteer. I'm a recruiter for a well-known EA org. I think I'm significantly more extroverted than the average EA. I don't have a technical background, so it's difficult for me to fully engage with some research-heavy EA content.

2. ^{^}

    Including another post about EAG Boston, already!! Never tell yourself that you'll quickly write a post about something on the EA Forum; someone will always post before you; theirs will be higher impact; you will be sad. On the other hand, the author of the other EAG Boston post completed 47 1:1s at EAG, which means I shouldn't be sad because he's in the top percentile of human-achievable productivity. Can't compare yourself to giants...

3. ^{^}

    As a descriptor of opinion ownership, not like I'm angry with anyone.

4. ^{^}

    Although, scarily for me, it apparently does represent GiveWell's heavy influence on its staff members' footnote usage. Anyway, if you're hungry for GiveWell opinion, it's just one click away.

5. ^{^}

    I'll say this once: Anything in this post that could possibly be interpreted as a joke instead of a serious claim/dig about cause prioritization is in fact a joke. AI safety has the top spot in my personal cause prioritization and I'd like to find a way to work on it more in the future. I also think animal welfare is important and I'll probably make jokes about that too.

6. ^{^}

    For example, I was able to schedule 1:1s, chat with other attendees, deal with stressful, brain-overloading notifications, read a few chapters of a book, complete a final exam for an edX class, listen to a few of my speakers' talks (Special shout out to the Rethink Priorities session on the CURVE Sequence, which I found quite enjoyable!), and just plain relax. At one point I even had a whispered 1:1 during a talk in the back of the room.

7. ^{^}

    I had a nightmare after the career fair in which a large group of attendees wearing "Just don't build AGI" shirts stopped at my booth, looked straight into my eyes, and said in calm and pleasant voices, "0% of our personal giving will be going to GiveWell's top-rated charities. Don't try to recruit us either; we will be seeking jobs in higher-impact cause areas than global health and development."

8. ^{^}

    You can fund a high-impact founder for $4k, think about it!

9. ^{^}

    Note: I'm not referring to the personal archetype of the same name.

10. ^{^}

     I kid. In reality, my one conversation about mechanistic interpretability at EAG lasted for <5 minutes. My interlocutor suddenly stopped talking and refused to say anything more about his work because he didn't know if he could trust me with dual-use research. Although I now know absolutely nothing about mechanistic interpretability, I gotta hand it to that guy's rock-solid infosec.

11. ^{^}

     In fact I'm going to EAGxVirtual in just a few weeks; you should too! If you have any questions about the event, you can ask them here.