SignalTech/AI
This new theory on how AI ‘thinks’ could impact whether it kills us all
This new theory on how AI ‘thinks’ could impact whether it kills us all
Context
What happened: An independent researcher, Nick Cammarata, who said he left OpenAI in 2023, published a report arguing AI neural networks are interpretable if examined closely, challenging the view that AI neurons are uninterpretable.
Key numbers: Cammarata said he left OpenAI in 2023.
Why it matters: AI analyst Andrew Curran said he had given up on real interpretability before superintelligence arrived and called the findings potentially big. OpenAI chief scientist Jakub Pachocki warned AI agents could obfuscate their visible reasoning.
What changed
An independent researcher, Nick Cammarata, who said he left OpenAI in 2023, published a report arguing AI neural networks are interpretable if examined closely, challenging the view that AI neurons are uninterpretable.
Why it matters
AI analyst Andrew Curran said he had given up on real interpretability before superintelligence arrived and called the findings potentially big. OpenAI chief scientist Jakub Pachocki warned AI agents could obfuscate their visible reasoning.