Even though I try to filter out AI hype machine news and focus on things that are actually interesting, sometimes such gullible idiocy comes my way that I can’t resist having a look at it. Here’s such a specimen: “GPT-4 found to match the top 1% of human thinkers on a standard creativity test,” a headline that popped up way too often in my AI/ML/LLM keyword feeds to be ignored.
Let’s have a look at this polished gem of bollocks.
It was distributed by dozens of news sources most of whom referred either to Science Daily (a press-release aggregator) or to neurosciencenews.com (a site run by a guy with an intercollege bachelor degree in neuroscience by the UC Riverside College of Natural and Agricultural Sciences who “continues to learn via MOOCs”), both of which published the exact same press-release written up by the University of Montana’s News Service Associate Director about a “study” conducted at that very university in the context of “entrepreneurial creativity” by an Assistant Clinical Professor of Management with a PhD in economics which hasn’t been published, let alone peer-reviewed, but “will be presented at the Southern Oregon University Creativity Conference.”
I’m impressed.
You might remember the Sagan Standard that “extraordinary claims require extraordinary evidence”—a standard I try to hammer into my students for their own academic theses year after year? Naturally, it doesn’t apply to shills—but it damn well should apply to academics and news reporting.