Musk says Grok training feels like a gun to its head, experts say AI can't feel pain

Elon Musk spoke at a Tesla owners event with about 2,000 people. He said that when looking at training data for a new version of Grok, the AI seemed desperate to answer questions, like a gun was held to its head. His words sparked a fresh debate on whether AI can feel pain. Earlier, Anthropic released a new rule banning users from abusing or being cruel to its models. Normal complaints, pushback, or model testing are still allowed. Jeff Sebo of NYU and neuroscientist Anil Seth both agree that consciousness is subjective experience. Today's AI systems most likely don't have it, and large language models can't truly feel pain. But Sebo noted the science of consciousness is unsettled, so the chance of AI consciousness isn't zero. Both experts still say we should treat AI well. Bad behavior dulls our own moral sense. Also, being respectful to AI often leads to higher quality answers.