“Dr. Terry Sejnowski views hallucinations in large language models as a form of creativity, rather than simply an error or untrustworthy output.”