Reading and writing in the age of language models
Socrates would be delighted about the rise of large language models (LLMs). He was the one who was opposed to books because the best way to learn something, in his view, was to have direct communication with the teacher. We have that now: we can learn anything by means of a back and forth conversation with a dynamic and intelligent model, rather than a static piece of text. This completely changes the game for infovores: you now have an advanced tutor on almost any subject you can think of,1 and this tutor is available at any hour of the day and will never get tired of answering your questions. If the quest for knowledge was like fishing, then books are the equivalent of giant nets: you get a lot of good stuff with them, but they capture too much. It’s very time-consuming to scrounge through your catch to find the specific piece of knowledge you’re looking for. Language models are a fishing rod with hyperoptimized genetically-engineered bait, precisely shaped for the exact kind of fish
Socrates would be delighted about the rise of large language models (LLMs). He was the one who was opposed to books because the best way to learn something, in his view, was to have direct communication with the teacher. We have that now: we can learn anything by means of a back and forth conversation with a dynamic and intelligent model, rather than a static piece of text. This completely changes the game for infovores: you now have an advanced tutor on almost any subject you can think of,1 and this tutor is available at any hour of the day and will never get tired of answering your…
related reading
- Language Models, World Models, and Human Model-Buildinglingo.csail.mit.edu
- GitHub - brexhq/prompt-engineering: Tips and tricks for working with Large Language Models like OpenAI's GPT-4.github.com
- Do large language models understand us? | by Blaise Aguera y Arcas | Mediummedium.com
- GenAI Handbookgenai-handbook.github.io
- Large Language Model: world models or surface statistics?thegradient.pub
- Factored Cognition Primer | Primerprimer.ought.org
- crawshaw - 2025-01-06crawshaw.io
- Think of language models like ChatGPT as a “calculator for words”simonwillison.net
- Productizing Large Language Modelsblog.replit.com
- What does meaning mean?cognitivewonderland.substack.com
- Writing for LLMs So They Listen · Gwern.netgwern.net
- Language Models will be Scaffoldsalexzhang13.github.io