← Back to context

Comment by aydyn

6 months ago

I think that perspective is mostly just romanticization of reading, which is fine; if you have a book reading fetish nobody is stopping you from reading books.

But in terms of efficiently grokking information I don't think it holds up.

Cliff notes is not the same as ChatGPT. Using cliff notes, you only have a fixed bullet point summarization. But with ChatGPT, if a bullet point is not sticking, you can ask ChatGPT to expound on it. Expound on the stories from Carnegie's life and how that lead him to the techniques he talks about. That mode of learning, having someone to bounce discussion back and forth is really efficient for me.

One problem is that depending on the book, the LLM can be somewhat cagey about if it really has read the book, or what sources it has used in constructing its “understanding” of it. Due to its sycophancy bias, the machine will laud your understanding of the text without a real referent to back it up, giving you what is essentially bullet point.

This may hack it for cursory knowledge, like how I know the rough outline of Ulysses, despite having never read it, but it cannot supplant a real textual reading of it. Despite all of this, LLMs can be great for comparing and contrasting writers and thinkers, and can help you organize your thoughts.