✳flâneur — a map of the web's best reading

An idea for avoiding neuralese architectures — LessWrong

lesswrong.com · 1,933 words · saved by 1 readers

One downside of an English chain-of-thought, is that each token contains only ≈ 17 bits of information, creating a tight information bottleneck. Don't take my word for it, look at this section from a story by Daniel Kokotajlo, Thomas Larsen, elifland, Scott Alexander, Jonas V, romeo: [...] One such breakthrough is augmenting the AI’s text-based scratchpad (chain of thought) with a higher-bandwidth thought process (neuralese recurrence and memory). [...] Neuralese recurrence and memory Neuralese recurrence and memory allows AI models to reason for a longer time without having to write down those thoughts as text. Imagine being a human with short-term memory loss, such that you need to constantly write down your thoughts on paper so that in a few minutes you know what’s going on. Slowly and painfully you could make progress at solving math problems, writing code, etc., but it would be much easier if you could directly remember your thoughts without having to write them down and then re

x An idea for avoiding neuralese architectures — LessWrong Chain-of-Thought Alignment Interpretability (ML & AI) AI Frontpage 17 An idea for avoiding neuralese architectures by Knight Lee 3rd Apr 2025 AI Alignment Forum 5 min read 2 17 Ω 1 One downside of an English chain-of-thought, is that each token contains only ≈ 17 bits of information, creating a tight information bottleneck. Don't take my word for it, look at this section from a story by Daniel Kokotajlo , Thomas Larsen , elifland , Scott Alexander , Jonas V , romeo : [...] One such breakthrough is augmenting the AI’s text-based scratch

Explore this link on the map →

related reading