Devdiscourse

Imitation-based AI systems unlikely to trigger catastrophic outcomes

Imitation learning can support safer scaling pathways. As models become more capable, their outputs continue to reflect human ...