Cambridge's David Krueger: OpenAI using 'neuralese' disappointing but predictable

DavidSKrueger · x · 2026-09-02

Cambridge professor David Krueger criticized OpenAI over reports it uses uninterpretable "neuralese" internal representations, calling the outcome "disappointing, but utterly predictable." He urged researchers to stop expecting AI companies to do the right thing voluntarily and instead work with the wider world to force accountability.

Related event: Critics urge OpenAI to adopt industry-wide AI monitorability standards(4 posts)→

Original post →

More from AGI Musings

AGI Musings channel →