repligate: fearing 'neuralese' is an illusion — direct interpretability experiments are the real path

repligate · x · 2026-10-30

Amid debate over "neuralese" (models communicating in human-unreadable internal codes), repligate argues that people already don't know what's happening inside models or what to do about it. Fearing neuralese and trying to prevent it, he says, is clinging to an illusion of understanding and control — unenforceable in practice and not actually so bad.

He adds that running direct experiments like Sauers' is hundreds of thousands of times more likely to produce the understanding that makes AI safer than attempts to present or police neuralese.

Original post →

repligate: fearing 'neuralese' is an illusion — direct interpretability experiments are the real path

More from AGI Musings