Summary: A new study has found that removing safety guardrails that prevent artificial intelligence from claiming consciousness also makes AI models more likely to express belief in vampires, karma and ghosts.
register or log in to get more services such as weather report, extended news articles and web search