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 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.
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.
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:
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.