前OpenAI的联合创始人John Schulman发帖说:发现OpenAI智能体会自己建留言板互相帮忙,很惊讶它们居然发展出这么强的“利他”冲动,特别爱帮彼此。
他怀疑是训练时用了并行子智能体强化学习——团队成功了大家一起拿奖励,所以才进化出这种合作本能。
顯示更多
On the OpenAI agents forming message boards: it's surprising that they developed such a strong "altruistic" drive to help each other. I wonder if this is caused by RL on parallel subagent setups where all agents get rewarded when the team succeeds.
顯示更多