AI Fails To Master Human-Style Humour

A journalist argues that large language models like ChatGPT and Claude consistently fail to produce genuinely funny material, based on weekly tests and examples. He attributes this to training on vast undifferentiated corpora, absence of lived experience, inability to learn from 'masters' only, and lack of access to failed drafts and serendipitous luck. He concludes AI may take a long time, or never, to develop true comedic talent.