I was told it would be a good idea to get people up to date with what I am doing and why via a weekly email. You can choose to read this or not, but I'm going to run the experiment anyway.
You may skip the transhumanist Silicon Valley science-fiction-insanity parts if you like. All the parts after "What I'm working on: small picture" will seem normal to you. I am a rather high-functioning optimistic and cheerful human, despite how weird and gloomy my beliefs may seem.
My normalness is partially thanks to you, dear friends. As far as research has shown, the most overwhelming factor in warding off burnout and depression is having good social relations.
What I believe:
The pace of AI development in conjunction with our utter lack of alignment capabilities makes human extinction likely within the next 2-5 years. This is a radical belief, even within the AI field itself (the average AI engineer only believes there's a ~ 10% chance of extinction), but I did not expect AI advances to happen so quickly. AI was supposed to be difficult: you were supposed to have to come up with a watertight definition of intelligence, and then come up with complicated algorithms to make your machine self-learn and then maybe after enough testing runs you could get something intelligent enough to beat a human at chess.
Nobody expected the solution to be as stupid as "dump incoherent amounts of data into a machine with a mind boggling amount of computing power". And yet here we are. The amount of computing used for AI has increased one hundred million fold since 2014. Sam Altman and the other heads of AI laboratories are explicitly aiming for an AI smarter than humans, sooner rather than later. Sam Altman wants to increase total computing power by another factor of 100 000 000. This does not bode well for safety, because the few safety efforts there are need more time to arrive at a solution.
This is the best explanation so far for why I believe all this. It's well written and simple, and from a simpler time, in 2016: https://www.lesswrong.com/posts/LTtNXM9shNM9AC2mp/superintelligence-faq
What I'm doing (big picture):
AI policy seems to be our best chance of survival. Most AI alignment organisations, which hoped to find a technical solution to the problem of encoding human values into an AI, are pivoting toward policy. From their perspective, any additional day we live and any increase in our odds--however slim--is a victory. This post explains it well, and it is how I am trying to go about things. I'm not familiar with the technical side of AI and can do more good in AI policy, which means getting the government to regulate AI in order to slow down AI capabilities.
This is possible. Almost all microchips are bottlenecked in Taiwan. The microchip supply line is centralized. Mostly, this is because a new chip factory costs more than 10 billion dollars in equipment alone, and R&D even more than that. Successfully restricting AI development would not be as difficult, policy-wise, as getting the US to enter World War II was. I'm an optimist.
The other option, besides policy, is getting rich fast and then dumping money onto people more competent than me. That other option might be more difficult than the first, but I'll take a shot at both just to make sure. (I suspect the startup world is more difficult to strive in than the political one, and trying to do both at once will be extra difficult.)
What I'm doing (small picture):
I'm running surveys on AI risk in the multiple campuses of Sciences Po, which is the foremost political institution in France. I hope that this will give me a lot of bang (getting powerful people to pay attention to AI) for my buck (time and effort). I have about 70 emails from their students, and that's who I send the surveys to.
I'm translating a paper on existential risk by an Oxford professor, Nick Bostrom, into French. It will be up on his website soon. The paper is this one.
I'm not doing much else. It makes me feel bad that I can't participate in bigger ways, and when I fail at completing these projects (or when it takes a few more days to make them), I feel guilty and terrible. It doesn't help that procrastination is directly linked, in my mind, with the odds of survival of my family and friends. If you can relate on some level, the blog post series replacing guilt is great.
Any advice or ideas would be appreciated. I have some ideas beyond what I have mentioned, but they are useless unless I execute them.
Practical discoveries:
I don't correspond to most measures of objective success, and most "self-help" advice is horrendous crap. This is not advice. It's what I've noted to seem to work for myself, in an entirely anecdotal way. If you want my advice, here it is: https://www.lesswrong.com/posts/gfbX3h4RueuDzWs7m/some-rules-for-life-v-0-0
Don't go to sleep after 1 AM. It's a terrible idea. I have enough data and anecdotal evidence to assert that even if it may seem like a good idea to continue working when it's early morning and you're enthusiastic, it is not. It screws over your immediate future self. It sabotages a day of your life. Sometimes I go to sleep so late I get a migraine that lasts for two days afterward.
Going to sleep at 1 AM regularly has been excellent though. I am just as conscious and capable as usual, if only slightly more tired at certain parts of the day. I used to go to sleep at 11:00. Two hours per day times 365 means I'm unlocking a full 30 days of consciousness every year. I get a lot more done.
Sleep patterns differ from people to people, so this might not be a good idea for you. But experimenting might not be. This short post is what convinced me to try it out: https://guzey.com/theses-on-sleep/
Always pay attention to the physiological explanation. Sometimes I feel quasi depressed during the day. I try to get work done in this state and then my mind becomes fragmented and I do something stupid like doomscroll on Twitter instead. It frustrates me, and then it's a vicious cycle where I get more and more fragmented and less and less able to work.
From the inside it feels like a black box: I don't know exactly where I fail to get things done, but at the end of the day, nothing is done. This is the worst feeling I experience daily.
Then I remember the circumstances: I went to sleep late last night so I'm tired, which correlates negatively with both agency (ability to go out into the world and do things) and happiness. Or, Im hungry because I only ate a little bit, and hunger is distracting and saps energy. So of course I'm more likely to feel depressed! Noticing the physiological origin of things has made some days a lot more bearable for me.
Focus on the utility. In economics, utility can be switched out for "what you value". I wrote in the last section about feeling like a black box when I procrastinate. I don't understand: all day, I've been making what seemed like perfectly reasonable choices! But somehow... No utility came out of the black box! Day by day goes by like this and now I feel like I've wasted a majority of my life from ages 13-16. I've wanted to start a YouTube channel for 3 years at least but somehow never managed to get this done. There was always a reasonable excuse: I need to find the fonts for my videos, or do this little research thing (it'll take 30 minutes I promise), or watch a few videos on how to start one first.
There's a script for what I wanted to be my first YouTube video that's been hanging in my docs for years, edited from time to time by past selves who thought they would give it a try. I've never published this video.
My first YouTube video had a script scribbled 5 minutes before and was deliberately filmed in the worst possible conditions I could find. If you've been procrastinating on a project, I encourage you to make it deliberately bad, in exchange for shipping it out by the end of the night.
The most important lesson I've learned for myself is to pay attention to the utility. That means observing yourself right now: what are you doing? Are you doing the thing you should be doing? Or something that supposedly serves that thing? Don't look at how the black box works, care only about what comes out.
I encourage you to write your own version of this, even if it's short (I wrote too much). I'd love to hear from you! You can reply to this email, even if it's just one sentence.
Good day,
Neil