Dwarkesh Patel: preference

Transcript context

…It's not just crystallized intelligence, but if you could see all the individual steps in my process, you'd be a lot less impressed. If you could see all of the times I just note down something like, “Hmm, that's funny.” Or, "Huh, another example of that," and if you just saw each particular step, you would say that what I was doing was reasonable and not some huge sign of brilliance. It would make sense to you in that moment. It's only when that happens over a decade, and you don't see the individual stuff, that my output at the end looks like magic. One of my favorite quotes about this process is from the magicians Penn & Teller. Teller says “magic is putting in more effort than any reasonable person would expect you to.” He tells a story about how they make cockroaches appear from a top hat. The trick is that they researched and found special cockroaches, and then found special styrofoam to trap the cockroaches, and arranged all that, for just a single trick. No reasonable person would do that, but they did because they wanted the trick to really pay off. The result is cockroaches somehow appearing from an empty hat. If you could see each step, it would make sense on its own, it would just look effortful. But when you see only the final trick, then that whole process and its output becomes magic. That’s one of the interesting things about your process. There are a couple of writers like Matt Levine or Byrne Hobart who write an article every day. I think of them almost like autoregressive models. For you, on some of the blog posts you can see the start date and end date that you list on your website of when you’ve been working on a piece. Sometimes it’s like 2009 to 2024. I feel like that’s much more like diffusion. You just keep iterating on the same image again and again. One of my favorite blog posts of yours is “Evolution as Backstop for RL,” where you talk about evolution as basically a mechanism to learn a better learning process. And that explains why corporations don’t improve over time but biological organisms do. I’m curious if you can walk me through the years that it took to write that. What was that process like, step by step? So the “Backstop” essay that you're referring to is the synthesis of seeing the same pattern show up again and again: a stupid, inefficient way of learning, which you use to learn something smarter, but where you still can’t get rid of the original one entirely. Sometimes examples would just connect to each other when I was thinking about this. Other times —when I started watching for this pattern—I would say, "Oh yes, ‘pain’ is a good example of this. Maybe this explains why we have pain in the very specific way that we have, when you can logically imagine other kinds of pain, and those other pains would be smarter, but nothing keeps them honest.” So you just chain them one by one, these individual examples of the pattern, and just keep clarifying the central idea as you go. Wittgenstein says that you can look at an idea from many directions and then go in spirals around it. In an essay like “Backstop,” it’s me spiraling around this idea of having many layers of “learning” all the way down.…

Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.