Hide table of contents

TLDR

I had a mix of feelings before and throughout EAG London 2024. Overall, the experience was excellent and I am more motivated and excited about my next steps in EA and AI safety. However, I am actually unsure if I will attend EAG next year, because I am yet to exhaust other means of networking, especially since I live in London.

Why might this be useful for you?

  • This is a narrative that is different to most others.
  • Depending on your background/personality, this will reduce the pressure to optimise every aspect of your time at EAG. I am not saying to do no optimization, but that there is a different balance for different people.
  • If you have not been to an EAG, this provides a flavour of the interactions and feelings - both positive and negative - that are possible.

My background

I did pure maths from undergraduate to PhD, then lectured maths for foundation year students for a few years, then moved to industry and have been a data scientist at Shell for three years. I took the GWWC pledge in 2014, but I had not actively engaged with the community or chosen a career based on EA principles.

A few years ago I made an effort to apply EA principles to my career. I worked through the 80000 Hours career template with AI safety being the obvious top choice, took the AI Safety Fundamentals course, applied to EAG London (and did not get accepted, which was reasonable), and also tried volunteering for SoGive for a couple of months. Ultimately the arguments for AI doom overwhelmed me and put me into defeatist mindset (‘How can you out-think a god-like super intelligence?’) so I just put my head in the sand instead of contributing.

In 2023, with ChatGPT and the prominence of AI, my motivation to contribute came back. I did take several actions, but spread out over several months:

  • I finally learned enough PyTorch to train my first CNN and RNN.
  • I attended an EA hackathon for software engineers and contributed to Stampy. The contributions were minimal though: shock-horror, the coding one does as a data scientist is not the same as what software engineers do!
  • I applied to some AI safety roles (Epoch AI Analyst, Quantum Leap founding learning engineer, Cohere AI Data Trainer)
  • I joined a Mech Interp Discord and within that a reading group for Mathematics for Machine Learning.

I go into these details to illustrate a key way I differ from the prototypical EA: I am not particularly agentic! Somebody more rational would have created more concrete plans, accountability systems, and explored more thoroughly the options and actions available. Despite being familiar with rationality / EA for several years, I had not absorbed the ideas enough to apply them in my life. I was a Bob who waits for opportunities to arise, and thus ends up making little progress.

The breakthrough came when I got accepted into ML4Good. I have written my thoughts on that experience, but the relevant thing is it gave me a huge boost in motivation and confidence to work on AI safety.

Preparing for EAG

I actually did not plan to attend EAG London! My next steps in AI Safety were clear (primarily upskilling by getting hands-on experience on projects) and I was unsure what I could bring to the table for other participants. However, three weeks before EAG, somebody in my ML4Good group chat asked who was going, so I figured I may as well apply and see what happens.

Given I am writing this, I was accepted! When reading the recommended EA Forum posts for EAG first-timers, I was taken aback by how practical and strategic these people were. This had a two-sided effect for me: it was intimidating and made me question how valuable I could be to other EAG participants, but it did also help me be more agentic and help me push myself outside my comfort zone.

In the end, I set these goals for the conference:

I organised around ten 1-1s: three with people in a similar situation to me, five in AI governance, one in AI safety field-building, one organiser of AI safety camp, and a couple not related to my goals but who had interesting non-standard profiles. I also planned to attend a couple of talks, a speed networking session, an ML4Good catch-up and office hours with Neel Nanda.

To my pleasant surprise, I also had a few people reaching out to me for 1-1s! I was not sure if I could help them, but naturally I accepted.

Positive experiences at EAG

I list these to show what you could get out of the conference. This is just my experience - there will be a huge range amongst all the participants.

General organisation and atmosphere

  • The networking was the easiest and most successful I have done. Reasons for this are:
    • We are all going into the 1-1s with clear goals, even if it is just to have an interesting chat.
    • EAs are more similar to me in conversational style than non-EAs.
    • Everybody is happy to help each other, making a supportive environment.
  • The conference was well organised and it is clear the team has learnt from previous runs.
  • I know there have been budget cuts, but as a first-timer, I did not feel that anything was missing.

My goals

  • I received constructive feedback and a variety of perspectives on the technical syllabus for non-technical AI safety people.
  • I learnt about new (to me) resources and organisations, e.g. Successif, Polaris, AI Safety and an AI safety volunteer program that Arcadia Impact is planning.
  • The organiser of AI Safety Camp suggested I could be a mentor for the next iteration! Given my limited experience, I had not considered this. Even better, I mentioned an idea for a project and they gave it a thumbs up!

Unexpected ways I can contribute to EAs

  • I am the only person on the EA forum to ask about tax deductibility of donations to EAG. One odd detail is that CEA said to donate via GWWC so GiftAid is applicable, but this suggestion is not made publicly. I decided to create a Manifold Market to estimate the value of EAG claiming Gift Aid on the donations.
  • I applied to LASR labs and I did not like the ‘ML coding test’, which turned out to be LeetCode problems. By chance, I had a 1-1 with the test-setter, and surprisingly they have no coding or ML experience. Furthermore, the exercises were checked by some mentors and they gave it a thumbs up. A positive is that the person was happy to hear my feedback. I will flesh this out in a separate post, because there are better ways to assess applicants.
  • A 3rd-year undergrad asked me for career advice in a 1-1. I provided my thoughts, but more importantly, I gave them the ‘obvious’ idea of asking 80,000 Hours for advice. They already knew about them, but somehow did not consider it.
  • The same undergrad planned a new habit of writing, with a goal of 60 minutes per day. I said that it is better to have small incremental goals to build a habit, so they changed it to 30 minutes. I pushed further, so they dropped it to 10 minutes.
  • I can promote the GWWC pledge and help dispel common misconceptions. Given my own path, I assumed most participants have taken the pledge, but the people at GWWC estimate it is around 30%.
  • Having conversations with EAs improved my rationality. I picked up the patterns, and am more likely to ask myself the important questions in the future.
  • Neel Nanda said the Deepmind Mech Interp team does not have any pressure to be profit-making. This is encouraging given the recent developments in OpenAI.
  • I learnt about AIM’s strategy, which is based on absorbency (how many people can a career path absorb) and provision of on-ramps for bottle-necks in EA career paths.
  • I learnt about Intelligence Rising’s RPGs, which highlight AI race dynamics. I asked what kind of lessons people learn and they recommended I read a session report.

Miscellaneous learnings, other

  • I was able to correct somebody’s incorrect belief that you need to start coding at the age of 11 to contribute to technical work, by giving myself as a counter-example: I was 29 when I started (properly) learning programming!
  • I learnt that the supplements with best evidence and effect sizes are Vitamin D (in particular Vitamin D3) and creatine. I also learnt that you should use zinc as *lozenges* and not tablets.

A bit of fun

  • I played Chess (managed to go from equal to dead lost in three moves: first blundering a pawn, then a rook, and then my queen…) and learnt a highly asymmetrical variant ‘Gravity Chess’.
  • On Sunday evening, I scooped up as much Huel as I could manage. I stuffed my backpack, put one 8-pack in a shoulder bag and carried two 8-packs in my hands. I ended up with 42 bottles…

Negative experiences at EAG

None of what I say should reflect badly on the individuals involved.

One person who asked me for a 1-1 was an 18-year-old about to start a maths degree. It turns out they know more about AI safety than me, e.g. asking about how the simulation model changes our reasoning about intelligent agents. I only have vague knowledge of this, so had nothing to say. Furthermore, the 18-year-old just got a full-time job as technical writer for Ars Technica, so they are about to start a maths degree while also having to write one article a day! Wild! To top it off, somebody else joined the conversation and I searched for them on SwapCard: they had an impressive profile, with a list of around 20 projects they have done or are interested in pursuing, in AI safety and GenAI. This admittedly put a dent in my confidence for a few hours.

My next 1-1 focussed on my idea of technical syllabus for non-technical people and I left with the feeling it was not a great idea. In retrospect, I cannot actually pinpoint why I left with this feeling, because I did leave with various practical and positive suggestions. Maybe I felt low because they were so able to quickly think of useful questions and ideas on the spot. Or maybe I was just not in a confident mindset from the previous conversation.

This general low feeling did stay with me for a few hours, but eventually subsided towards the evening. On Sunday, I was actually so exhausted that my mind did not have enough energy to go into negative spirals (exhausted because of non-EAG reasons: I stayed up until 4am at a close friend’s party. Not recommended.)  But I did still have one negative bout.

At lunchtime on Sunday, I saw a Facebook acquaintance I had never met, so I joined them for lunch. They were already in a conversation about the risk of civil conflict in the USA and its consequences on AI safety: would this slow down AI capabilities, would AI labs send staff to a safer country, would the US government nationalise the AI labs to prevent staff from leaving, etc. I did not have anything to add, so I quietly ate my lunch while listening to them bounce ideas off each other. This does not make me feel terrible - it is normal for me to not contribute in group conversations - but for those 15 minutes or so, I felt inadequate to an extent.

Why I am unsure about attending next year

The positive experiences outweigh the negative significantly, so why am I unsure of attending? It boils down to two things:

  • The cost of the event feels high and my gut instinct is that I did not get £1000-1500 worth of value. For comparison, ML4Good has similar cost per person and for me it had a greater impact.
  • I have not exhausted the other ways I can network with the community, especially given that I live in London. But even if I did not live in London, there are plenty of (free) opportunities online, e.g. Apart hackathons.

This is not a criticism of CEA and I know there are good reasons why the event is expensive. I am likely under-valuing the benefits I received and over-estimating how replicable the benefits are online.

Recommendations

  • Advice from elsewhere that I repeat: set goals for the conference, focus on 1-1s, plan ways to relax
  • Advice I have not seen emphasised: try speed networking sessions, be aware that people are highly confident and you will find people who are equally confident with opposite views.
  • Read Amber Dawn’s exploration on how EA conferences can be emotionally difficult. My experiences match with the ‘comparing to others negatively’ category, but there are several others.

Thanks to Emily Fan and Matthias Endres for reading a draft and their helpful suggestions.

58

0
0

Reactions

0
0

More posts like this

Comments2
Sorted by Click to highlight new comments since:

Thanks for writing up your experiences here; I found them interesting to read/reflect on!

A 3rd-year undergrad asked me for career advice in a 1-1. I provided my thoughts, but more importantly, I gave them the ‘obvious’ idea of asking 80,000 Hours for advice. They already knew about them, but somehow did not consider it.

Woo thanks so much for doing this! My impression is that a lot of people just need this kind of nudge :)

I'd have benefited from that kind of nudge myself! I was aware of 80K for years but never even considered coaching.

Curated and popular this week
Relevant opportunities