I apologise to all of you who have to scroll through yet another person reminiscing about sitting in one place for 30 days writing about things nobody cares about, but I have official word from resident Frenchwoman Lucie Philippon that "Inkhaven retrospectives are not inkslop," so here we go.
I arrived at Inkhaven at 11pm PT on the 31st of March, 8am on the 1st of April CET, after a delay at immigration meant I missed my connecting flight from Las Vegas[1].
Bright and early the following morning, I got to work on my first post, "General intelligence is hitting a wall", which is to date my most upvoted post on LessWrong and (in the opinion of both me and Claude) my best post of Inkhaven. I guess I clearly haven't improved since then.
I posted it shortly before 10pm, having sent it to Justis, who cleared out the worst of the mistakes I was making.
The following couple of days were some of my less good posts – I wasn't yet used to the pace of writing, and I was spending all my time getting to know my new friends.
I think a broad pattern I found was that the most valuable aspect of Inkhaven was the conversations I had, with fascinating insights happening constantly. I was completely unable to keep myself away from them, and my resulting output was crap. I found that because I had to publish so fast, despite the posts being ideas I'd been working up and thinking about for years[2], when they came out of my fingers, they came out like garbage. Oh well, guess I'll know for next time.
Posts on April 6 and 7 were much improved. Contra Nina Panickssery on children was a response to a highly upvoted post I thought was mostly wrong. Most people can't juggle one ball was a guide to juggling I wrote because I'd been teaching people to juggle, and Valentino from SMTM suggested I should write a post about it. It got to the front page of Hacker News. It's funny that where I thought at the time the previous posts had been grand insights, this was just obvious stuff I pooped out easily (because it was obvious), and everyone loved it. Note to self: Write down the obvious things you regularly tell people.
On April 8th and 9th, I published "Stockfish is not a chess superintelligence (and it doesn't need to be)" and "AI identity is not tied to its model", which I think people found somewhat interesting.
On April 10th I published the somewhat controversial "Getting Claude to rank the Inkhaven bloggers", which later earned me the no. 1 spot on "Which 14 residents should be kicked out of Inkhaven?".
On April 11th, in response to an escalating debate about veganism in the Inkhaven community, I published "Eating meat is fine if you live in a simulation", the funniest part of which is the title.
On April 15th, I published "How to run from a bull", an account of my participation in the Running of the Bulls in Pamplona. This was entirely not the sort of writing I wanted to be doing, but posting about things that you've done is actually quite a lot easier than posting about complex topics you have to think carefully about how to explain.
I had a wonderful meeting with Justis Mills, who read through all my posts and basically recommended that I post travel stuff so that I had more time to write the interesting AI takes he thought were my best posts. I agreed and promptly wrote travel posts for the rest of Inkhaven: "A long hike", Parts 1-10, about walking 500 km from Ho Chi Minh City to Siem Reap, and "Playing temple run irl", Parts 1-3. They're fun posts and probably improved my writing, but they also were not what I came here for. Oh well.
Despite the writing not being all that I wanted, the social more than made up for it. I met a tonne of amazing people, whom I miss dearly; I left on the 26th of April to fly to Czechia to start the AFFINE seminar. It will be interesting to see how they compare.
I just read Henry Stanley's piece, and he's given advice better than I can, but here are a few reflections from my time anyway:
- I was doing BlueDot facilitation and AI Safety Camp online at the same time as Inkhaven. Don't be like me; this sucked[3].
- Do have a list of things you want to write about before coming in.
- Don't expect to follow it.
- Do go outside the compound (I did so like 3 times, and it was worth it each time).
- Do spend a bunch of time chatting to whoever is around the Aumann kitchen (everyone is interesting)
- Do karaoke. It was awesome, and I have massive fomo that I'm in Europe and everyone is currently karaokeing at the afterparty at Lighthaven without me as I'm typing this.
- Go to the winners' lounge when you've posted. Some amazing discussions were had there, both in terms of entertainment and insight. Anna Mattinger has posted at length about one of them.
- Find the Sisyphus table. It's really cool.
- Get advice from the writing advisors. Like, seriously, they're right there and they're really helpful.
- ^
The customs officer was, it seemed, trying to trick me into contradicting myself. He asked me each question 3 times in different ways. Since every customs officer was doing this for everyone, the whole queue took forever despite my flight connection winning me a queue skip.
- ^
Before Inkhaven, I fully planned out which posts I was going to write on each day, going so far as to get Claude to rank my post ideas using an Elo system from the outputs of my comparisons. I followed it for a week.
- ^
To be clear, I think that these are both worthwhile by themselves, but balancing these commitments with Inkhaven made it much harder to concentrate on what I was doing. This is probably also why some of my stuff was bad.
Discuss