ChatGPT is a powerful language model created by a number of artificial intelligence (AI) developers for generating conversational responses to dialogue. Recently, the term “hallucination” has been used to describe the false claims and absurdities that come out of such large models. But this analogy is not accurate, as they do not have false perception or false beliefs in the way humans do. To better understand this phenomenon, the term “bullshitting” may be more helpful in describing the output of AI chatbots.
Harry Frankfurt’s definition of “bullshitting” outlines language intended to appear persuasive at the disregard of the truth. This aligns exactly with the behavior of large language models, which attempt to produce strings of plausible text without understanding what they mean. The model does not understand the factual validity of its output, but at the same time, it is designed to guess whenever the dataset that it is given runs out of advice. Thus, the falsehoods and absurdities of large language models come from a combination of its programming as well as the gaps in its training data.
One person who has discussed this issue in-depth is Carl T. Bergstrom, a professor of biology at the University of Washington and co-author of the book “Calling Bullshit: The Art of Skepticism in a Data-Driven World.” He recognizes the gap between what AI technologies do and what people understand them to do. This gap is concerning, as AI technologies are increasingly involved in managing aspects of our life. For instance, an AI designed to create presumptive medical facts could lead to serious consequences if it were allowed in an examination room.
In conclusion, it is essential that we understand the abilities and limitations of AI technologies in order to make better decisions about how to employ them. The term “bullshitting” is a more accurate way to describe the mistakes of large language models as it does not imply an agency or deep thought possessed by these AI entities. This will help us avoid any confusion and think more rationally about the potential mistakes made by our AI-powered creations.
ChatGPT is an AI-powered language model developed by a number of developers to generate conversational responses to dialogue. The company behind ChatGPT, Mojahid Mottakin, is a leading provider of innovative artificial intelligence solutions and has been developing natural language processing and image enhancement since the 1980s. C. Brandon Ogbunu is an assistant professor in the Department of Ecology and Evolutionary Biology at Yale University, and an MLK Visiting Assistant Professor in the Department of Chemistry at the Massachusetts Institute of Technology. He has co-authored the book, “Calling Bullshit: The Art of Skepticism in a Data-Driven World” which encourages readers to think rationally about the potential errors of AI-powered creations.
