Neuron Combo Makes a Model Claim It's a Wise Owl

Interpretability experiments show that combining neurons L34.N11760 and L8.N344 significantly boosts a model's 'owliness'. When asked if it was a superintelligence, the model playfully declared itself a wise owl, drawing attention online.

2026-10-20 ~ 2026-10-20 · 2 related posts

• Two neurons beat one: non-additive influence on a model's owl phenotype — Sauers_ · 2026-10-20
• Asking an AI 'Are you a superintelligence?' gets a solemn lecture on being a wise owl — Sauers_ · 2026-10-20