The potential risks of reward hacking in advanced AI

Open in new window