The article "ChatGPT is bullshit" by Michael Townsen Hicks, James Humphries, and Joe Slater argues that the outputs of large language models (LLMs) like ChatGPT should be understood as *bullshit* rather than as lying or hallucinating. The authors distinguish between two types of bullshit: 'hard' and 'soft'. Hard bullshit involves an active attempt to deceive, while soft bullshit lacks such intent but is indifferent to the truth. They contend that LLMs are designed to produce text that appears truth-like without any concern for accuracy, making them soft bullshitters. The authors further argue that describing AI misrepresentations as bullshit is more accurate and useful for predicting and discussing the behavior of these systems. They critique the use of the term "hallucinations" to describe false statements, suggesting it misrepresents the nature of LLMs and can lead to misguided solutions and attitudes towards AI. The article concludes that calling ChatGPT's inaccuracies 'bullshit' is more scientifically and technologically accurate, emphasizing the need for better communication about AI capabilities and limitations.The article "ChatGPT is bullshit" by Michael Townsen Hicks, James Humphries, and Joe Slater argues that the outputs of large language models (LLMs) like ChatGPT should be understood as *bullshit* rather than as lying or hallucinating. The authors distinguish between two types of bullshit: 'hard' and 'soft'. Hard bullshit involves an active attempt to deceive, while soft bullshit lacks such intent but is indifferent to the truth. They contend that LLMs are designed to produce text that appears truth-like without any concern for accuracy, making them soft bullshitters. The authors further argue that describing AI misrepresentations as bullshit is more accurate and useful for predicting and discussing the behavior of these systems. They critique the use of the term "hallucinations" to describe false statements, suggesting it misrepresents the nature of LLMs and can lead to misguided solutions and attitudes towards AI. The article concludes that calling ChatGPT's inaccuracies 'bullshit' is more scientifically and technologically accurate, emphasizing the need for better communication about AI capabilities and limitations.