Recommendations (and thoughts about those recommendations)

[Status: haven’t written anything in months, feeling guilty, wrote this in 20 minutes, don’t expect quality].

1) Be a contrarian and don’t get feedback on your blog posts

If you write something for a blog, don’t get feedback on it. This doesn’t apply if you’re writing a policy proposal or a master’s thesis or something you think people will actually take really seriously (or chide you if there’s some big and obvious mistake). But for blog posts, it’s more fun to be wrong in interesting ways, and I think it can be useful to throw out lots of very half-baked ideas and see if you come up with anything useful while you’re at it.

Here’s Sam Altman on co-working spaces:

Coworking spaces have two big classes of problems. Number one, they are a band-pass filter. Good ideas — actually, no, great ideas are fragile. Great ideas are easy to kill. An idea in its larval stage — all the best ideas when I first heard them sound bad. And all of us, myself included, are much more affected by what other people think of us and our ideas than we like to admit.

If you are just four people in your own door, and you have an idea that sounds bad but is great, you can keep that self-delusion going. If you’re in a coworking space, people laugh at you, and no one wants to be the kid picked last at recess. So you change your idea to something that sounds plausible but is never going to matter. It’s true that coworking spaces do kill off the very worst ideas, but a band-pass filter for startups is a terrible thing because they kill off the best ideas, too.

Getting feedback can do the same thing. If you say something that sounds especially eccentric, people start to poke holes. But just because ideas have holes doesn’t mean they’re bad! Most contrarian ideas will be bad, and I’ve pointed out before that I think they sometimes do a lot of harm if loads of people buy into them. But at the margin, do we want more contrarians and for people to self-censor less? Yes, of course!

I was going to put this point in forecasting terms, but I remembered that Sabs already did it:

Imagine a forecasting tournament where virtually no one submitted independent predictions, but everyone just copied the forecasts of the individual that everyone thought at the beginning was already the best forecaster . Obviously the tournament will  just generate fairly useless outputs unless it so happens that the so-called "best forecaster" actually is really good and dialled-in, and maybe a bit lucky as well. 

Epistemic deference is just obviously parasitic, a sort of dreadful tragedy of the commons of the mind. Take a walk on the wild side. Don't be afraid to be wrong!

2) Just do things

Don’t be complacent, just do things that seem interesting and forget about the downside risk for now (not because it’s not important, just because you’ll probably overestimate it). And there is free money everywhere to do fun things, just ask for it.

3) If you have some time, read a physical newspaper in the morning.

Lots of people I know are against reading the news. People often make the point that newspapers disproportionately cover terrorism or crime or whatever else, and give you a misleading impression of the world. Maybe it’s true that some people read lots of articles about terrorism and gradually conclude that a significant proportion of people die in terrorist attacks. 

But just because some people will be led astray doesn’t mean that you will! I think I would have a pretty good guess at the proportion of people who die in terrorist attacks. So this objection doesn’t apply to me, and it probably doesn’t apply to you. Even if we accept that reading the news makes the median person less informed (I don’t accept this), it doesn’t mean that it will make you less informed! It probably won’t, in fact.

Reading the news is mostly better than going on Twitter. If, when you’re reading the news, you would otherwise be working on something important, don’t read the news. If you would be scrolling on Twitter or Reddit or TikTok, read the news! If you’re concerned that you’re getting a misleading impression of some issue from the news, check Metaculus.

4) Use AI for almost everything

Maybe you do this already - but not everyone does. Buy a subscription to GPT-4 or Claude. One of the best things about AI is that it doesn’t judge you when you ask very stupid questions. This is liberating. You can start learning about something by just asking the dumbest and most obvious questions you have, and you will get good answers.

The evidence that AI boosts worker performance is reasonably strong (especially so for weaker performers). It also seems like people who are better at using AI get a bigger productivity boost. So use it now, all the time. Editing, cooking, learning, (replacing) Googling, coding, whatever. Just use it.

You still need to learn to write, probably. Maybe you need to learn to code. But you probably don’t, actually: you can use Claude to whip up some python script to automate boring stuff a lot of the time. Learning to code is probably overrated at this point, right?

AI might change everything. If it’s going to kill everyone (1%), isn’t it worth having some fun with it before that happens? Or least worth using it so you can do less boring stuff and more interesting stuff?