If you want to fool Pangram, you'll need better arguments. That's the takeaway from an interview with Max Spero, CEO of AI text detector Pangram, published on AI Policy Perspectives.

Spero calls Pangram's deep-learning classifier a black box. "We don't have a ton of interpretability into why it makes the predictions that it does," he said. The tool surfaces suspicious phrases as clues, but the model picks up on structural patterns a language model leaves behind when organizing a document. Even Pangram doesn't fully understand those patterns.

Spero also argues that language models "might be" better than average humans at grammar and logic but are far more uniform. Ask an LLM for 100 arguments on a topic and they'll cluster in a narrow band, "whereas the space of human arguments is going to be very diverse."