I'm surprised that a large company like OpenAI halted its biggest planned reinforcement‑learning project. There will be many safety problems and risks. How could they not have predicted this, given their funding and scale?