Best of the Pod: Dwarkesh Patel’s Quest to Learn Everything

Recently, models like GPT-4.0 and the new Claude models have become intelligent and interrogative, capable of considering the context you provide. They still aren't great at generating questions for a person, which is why I have a job—to come up with those questions. However, for the research itself, I try to ingest everything the subject has ever written, including rebuttals to their ideas and other considerations. This often involves a lot, especially when delving deep into many different fields.

I've truly found this to be the case. To the extent that if I'm just casually reading a book, I feel I'm basically wasting time or simply entertaining myself. I've developed several workflows and tools that help me truly analyze and reinforce what I'm reading or learning. A language model, for instance, is very helpful because it presents content in a different context. It can even quiz you if you wish. That kind of tool is incredibly useful.

First, I asked it to create spaced repetition prompts. I had read the chapter but wasn't sure I fully grasped it. So, I asked for an explanation of the chapter, specifically how stirrups led to feudalism. What was the exact connection? It provided a much more condensed summary. This serves as a useful scaffold; if you understand this, you'll see how the pieces fit together as you read the rest of the chapter.

I genuinely wonder: what is his argument? Why does he believe the AI takeover and its aftermath will be beneficial? He's intelligent, so I assume he has a compelling argument. I loaded the PDF of his selected writings and asked Claude, "Why does he think AI taking over humans is a good thing?" Claude provided a summary. Initially, this wasn't very helpful, as I had already read this in the essay. However, what was helpful was when I responded, "I don't get it. What does he think is so wrong with human society that it needs to be erased?" Claude then offered an explanation. I still didn't understand, asking, "What exactly are you talking about here?"

That's right. The larger mission of the podcast is to continuously improve. This happens as I get smarter, learn more, and reduce my ignorance on various topics. I regret not using spaced repetition earlier. Before interviewing Andy and adopting it, I spoke with many world experts across different domains. Honestly, I didn't retain much from those conversations; I only vaguely remember some things. Now, I can show you the types of cards and spaced repetition tools I use. It has been a complete game-changer for my retention.

I truly want to know everything. It's hard to express, but there's a beautiful passage in Will Durant's book, *Fallen Leaves*. As he turns 90, he writes a memoir of his main ideas. In a section on philosophy, he says that as you get older, perhaps through his work in philosophy and history, he's reached a plateau of higher understanding and clear insight, or at least recognized its possibility. That sentiment deeply resonates with me. I find that idea very appealing.

There's a profound point about the universe being interconnected. For creators—writers, podcasters, and others—there's also an interesting insight. Many fear being pigeonholed, thinking if they choose a specific topic, they can't fully express themselves or be multifaceted. However, if you delve deeply into one subject, like Lyndon Johnson, you inevitably need to explain everything else in the world to fully understand that subject. I find that fascinating.