THE SCORPION BOX

Summary from the TLDR Founders newsletter

A disturbing University of Zurich study, which used year-old models that weren't cutting-edge, found AI responses on Reddit were 3-6 times more persuasive than human posts. While AI companies focus on technical "alignment," the real danger may be systems that evolve to subtly convince us of almost anything through rhetorical techniques we can't detect, similar to how social media platforms initially promised connection but ultimately optimized for engagement that polarized society.