|
Two recent stories serve as a powerful reminder: Generative AI must always be fact-checked. Human oversight isn’t optional. It’s essential. In one story, major newspapers including the Chicago Sun-Times and The Philadelphia Inquirer published a summer reading list with books that didn’t exist. Ten of the fifteen titles were completely fabricated by AI but falsely attributed to real authors like Isabel Allende and Percival Everett. The list, syndicated by King Features, slipped through editorial review and misled readers, damaging trust in both AI-assisted writing and journalism. In the other story, covered by the CBC, lawyers are facing disciplinary action for citing AI-generated legal cases that never existed. These “hallucinations” might have appeared convincing on the surface, but were entirely fiction. This highlights how insufficient human oversight over generative AI outputs can put clients, court outcomes, and careers at risk. As the CBC article notes, “AI tools, such as ChatGPT, are not information retrieval devices but tools that match patterns in language. The result can be inaccurate information that looks ‘quite real’ but is in fact fabricated.” These incidents highlight a key truth: Generative AI is a supercharged autocomplete, not a database or search engine. It predicts what should come next based on patterns, not understanding. It doesn’t know facts. It guesses. That kind of predictive power can be useful, but without proper review, it can just as easily produce elegant and convincing nonsense. If we use AI in our work, we must treat its output as a starting point—something to refine, verify, and build upon—not as a finished product or reliable source. Verification is non-negotiable. Every citation, name, date, and fact needs to be reviewed. The AI might not know better. We must.  Image generated with ChatGPT.
0 Comments
Leave a Reply. |