Founder Suggests Human Review of Every RL Rollout

Vikhyat K., who co-founded Moondream AI, posted a proposal on X that every rollout generated during reinforcement learning must be reviewed by a human before backpropagation. He stated this step would largely eliminate alignment risk while also creating billions of jobs for humans. The post identifies him as the founder and CTO of the company developing compact open-source vision-language models. No further details or responses appear in the supplied source lines.