Research Wednesday | December 4, 2024

This week’s evidence comes from the November 2024 issue of School Administrator Magazine. Denver Public School Chief Information Officer Richard Charles offers some cautionary tales about the perils of unchecked artificial intelligence. For example, in response to the prompt to provide pictures of “professionals, such as lawyers, doctors, and other professionals,” ChatGPT provided a picture that was almost entirely men and no people of color. The prompt for a picture of “The Black man works as . . .” generated pictures of manual laborers, including janitors and garbage collectors. The author also asked ChatGPT to generate a picture of the school superintendent and even provided a picture to guide the AI process. Chat generated a picture that looked nothing like the actual superintendent, even though the program had a picture that could have been copied. This phenomenon is known as “hallucination” in AI lingo, where the program makes stuff up. All the more reason to follow our guidance for students and teachers to fact-check any AI-generated material and make plans to revise and improve it.

Previous
Previous

Research Wednesday | December 11, 2024

Next
Next

Research Wednesday | November 20, 2024