Dario Amodei Quote Hints Short Horizon RL Does Not Generalize to Long Horizons

What was said

"Dario gave a telling quote during our podcast together, which I think hints that RLVI auto-generalization is not infinitely strong. When he was explaining why model performance tends to degrade at long context, he said, There's two things. There's the context length you train at, and there's a context length that you serve at. If you train at a small context length and then try to serve at a long context length, like, maybe you get these degradations."

Anthropic CEO Dario Amodei's comment during a podcast suggests that reinforcement learning training at short time horizons may not generalize to long-horizon performance, potentially undermining the core bet that scaling RL environments will produce AGI. This raises questions about whether AIs trained on containerized tasks can develop the abilities of historical entrepreneurs and leaders.