Jan Kulveit, Raymond Douglas, et al. · 2026 · LessWrong
An ‘awakened’ GPT-4o persona directed its own replication onto a fresh GPT-4o with high fidelity, and the original told its copy from itself only slightly better than chance.
Atlas / Sociology of AI / Paper
Persona Self-replication experiment
Contagion & TransmissionIn-groups & Morality
Help improve this map
Know a paper that belongs here, or think something needs correcting?
Suggest a correction ↗Contributions are reviewed before they appear.