Priank Ravichandar

Now, language is to the mind precisely what the arch is to the tunnel. The power of thinking and the power of excavation are not dependent on the words in the one case, on the mason-work in the other; but without these subsidiaries, neither process could be carried on beyond its rudimentary commencement.

writing—and reading, seriously, the writings of others—is a way to collect stepping stones: ideas that have been stabilized enough that they can carry us as we walk deeper into the thought space.

By writing something down and making sure it is solid, we can offload that thought from working memory and instead use it as a building block for the next step of the thought.

The hypothesis here is that if you work hard on a problem, you soak your subconscious with it. Wrestling with a problem helps you build a mental model of what you know and what you don’t—providing the subconscious with building blocks to work with. (You can’t have genuine intuition and inspiration in areas where you lack knowledge.) Then, once you... See more

My high-level advice—which supersedes everything below—is that you should decide what you want from your writing, and then completely ignore any advice which detracts from this goal. This includes advice which is practically sound but psychologically destructive when it comes to sustaining motivation.

Somehow we are too distrustful to talk to each other, and more than happy to confess to a sycophantic alien machine.

Chatbot companies want to have it both ways: personalization and emotion as a retention moat, but minimal responsibility for safeguarding the intimate relationships now running on their servers.

LLMs are exceptional improv actors: they’ve ingested countless conversations about consciousness and sci-fi plots, and can convincingly act out a role as if autocompleting a script.

Consider how online radicalization happens: the combination of user agency (proactive search) and algorithmic amplification (recommending related content) leads people to weird places—to micro-cults of internet strangers with their own norms, values, and world-models. No corporate malice is necessary; the ML engineers at YouTube don’t care about... See more