if less wrong is not to be a true competitor to arxiv because of the difference between them in intellectual precision^1 then that matches my intuition of what less wrong should be much better: it’s a place where you can go to have useful arguments, where disagreements in concrete binding of words can be resolved well enough to discuss hard things clearly-ish in English^2, and where you can go to future out how to be less wrong interactively. it’s also got a bunch of old posts, many of which can be improved on and turned into papers, though usually the first step is going to be literature search to link ideas back to their academic traditions, then write out the math in the syntax of the tradition you want to publish in. to put it another way, less wrong is a self-teacher’s academic institution; but as such you need to come into it expecting it to be a game of collaborative self improvement where you and others will both be often wrong, often briefly confidently wrong, occasionally enduringly confidently wrong, and you want to figure out which direction is less even with all the different kinds of errors you see.
because of the high rate of memory nonretention from single examples like a blog post, I think we should be much more willing to restate ideas when in brainstorming shortforms, because loading things into working set is good and helps both humans and LLMs load context.
because of highly exploratory nature, explorations should likely be short to engage others’ exploratory input.
commentary appreciated, venue redirect welcome.
ft1. see recent discussions about publishing on arxiv
ft2. or perhaps other languages
if less wrong is not to be a true competitor to arxiv because of the difference between them in intellectual precision^1 then that matches my intuition of what less wrong should be much better: it’s a place where you can go to have useful arguments, where disagreements in concrete binding of words can be resolved well enough to discuss hard things clearly-ish in English^2, and where you can go to future out how to be less wrong interactively. it’s also got a bunch of old posts, many of which can be improved on and turned into papers, though usually the first step is going to be literature search to link ideas back to their academic traditions, then write out the math in the syntax of the tradition you want to publish in. to put it another way, less wrong is a self-teacher’s academic institution; but as such you need to come into it expecting it to be a game of collaborative self improvement where you and others will both be often wrong, often briefly confidently wrong, occasionally enduringly confidently wrong, and you want to figure out which direction is less even with all the different kinds of errors you see.
because of the high rate of memory nonretention from single examples like a blog post, I think we should be much more willing to restate ideas when in brainstorming shortforms, because loading things into working set is good and helps both humans and LLMs load context.
because of highly exploratory nature, explorations should likely be short to engage others’ exploratory input.
commentary appreciated, venue redirect welcome.
ft1. see recent discussions about publishing on arxiv ft2. or perhaps other languages