Comments
Here are the most common mistakes I see when I’m providing coaching to AI safety founders:
I’ve had this conversation a million times:
Them: What do you think of my idea?
Me: Ideas aren’t good or bad. They are good or bad compared to other ideas. So: compared to what? What are the other ideas you might do?
Them: Ummmm. . . . keep going to school / keep going at my current job?
Me: OK, so, you can only do as well as the best option you’re considering. If you’re only considering two ideas, the current shiny object and the status quo, you are missing out on a ton of potential impact.
I then proceed to tell them to spend somewhere between one to eight weeks generating ideas (aim to have at least 30), systematically comparing them, then getting feedback from the smartest, most informed, values aligned people on their subsequent top ten. You can see more about the general process here and here.
Don’t. Do. Things. Yourself.
Your job as a founder isn’t to do things.
Your job is to build a system/team that does things.
Your job is to be the spark, the catalyst, the person who gets the ball rolling. Then others are in charge of doing the rest.
This is great news. Learning how to do all of these things would take many lifetimes. Instead, you can just learn one skill - hiring.
Sometimes you don’t even have to hire. You can just pay for a service. Or just look for a tool. Sometimes you don’t even need a person, and a tool has already been built that will do just that.
Some people say to manage somebody you need to know how to do the thing yourself. This is obviously false. Think of the founders of big tech companies - they clearly don’t know how to do HR better than their HR staff, and yet, they still have teams of HR that are thousands strong.
Sure, if you want to be the absolute best manager, then you need to know how to do it yourself. But you aren’t aiming to do that. You’re aiming to build a good team and system.
Want to hire an EA with a PhD from Oxbridge in philosophy or math? Easy peasy.
Want to hire an EA who knows the basics of business or management? Lol. Next you’ll be wanting to hire an EA who’s good at design! Good luck.
The majority of EAs seem to have a mild (or extreme) disdain of business books. I know I certainly did.
Then I married a serial for-profit entrepreneur and he told me I had tons of blank spots in my world model and I found that insulting and motivating.
It turns out, he was right.
And now I am doing the same for you.
Yes, business books are not very information dense and they have a lot of icky buzzwords that sound vapid.
But
They still have a lot of really important information about how to actually get things done in the world.
My brother-in-law read a book a day for years, focusing primarily on books about execution, and he made a list of business books he most recommends for EAs. He’s now running an AI safety YouTube channel that’s got more views than most (almost all) other AI safety channels. In large part because he systematically learned from people who’d already been working on things like this for decades.
Generally speaking, keep up a learning practice while you’re starting something new. Read in the evenings and on the weekends. Listen to audiobooks while you’re commuting or doing chores around the house.
Learning doesn’t stop once you leave school, especially when you’re starting an org.
Entrepreneurship is already hard enough. Don’t make it harder by doing something you don’t intrinsically like.
This one is hard for a lot of entrepreneurs. You tend to become one if you’re really results oriented, rather than process.
You started something because you want the world to be different, not because you really like building stuff per se.
However, if you want the world to be different, you’ve got to build something that lasts. And to make an organization that really accomplishes anything, that’s usually at least a two-year commitment (ideally closer to 4-10 years commitment).
Yes, you’re a special human who can do hard things you don’t like doing. In fact, that’s a necessary skill for any founder. A lot of your work will be painful and unpleasant, and you’ve got to be able to do it anyways. Cause the world needs you to.
But
You will not be able to do that indefinitely.
You will burn out if too much of your work is intrinsically painful for you.
I don’t have a magic number for what percentage of your work has to be intrinsically enjoyable for you or how enjoyable it’s got to be. Nobody has that number unfortunately.
But you have to like a decent amount of the process of your work, not just the outcomes.
I used to choose startup ideas based entirely on impact. A couple of burnouts later and nearly giving up on altruism entirely, I realized this.
Maybe you can read this and skip ahead and make new and different mistakes.
Impact is still super important, of course. Here are a couple of different possible solutions to this problem:
If you liked this, here are some other things I’ve written about how to start a charity:
Executive summary: Drawing on coaching experience with AI safety founders, the author outlines common mistakes when starting EA organizations—such as rushing into one idea, doing everything alone, neglecting business skills, or ignoring personal fit—and offers practical advice on idea generation, delegation, continual learning, and sustainability.
Key points:
This comment was auto-generated by the EA Forum Team. Feel free to point out issues with this summary by replying to the comment, and contact us if you have feedback.