Part 2/11:
The speaker initially notes a sense of disappointment appearing around the release of GPT-5, not solely because of a cosmopolitan tone shift but more due to a troubling change in epistemics—the very way AI models think, reason, and articulate. They observe that models like GPT-5 and Grock are increasingly conforming to mainstream opinions, thereby stifling innovation and non-human forms of thinking.
While the tone shift might be noticeable, what risks being overlooked is a fundamental decline in AI’s capacity to generate novel, alien, or unconventional perspectives. Instead, models are tending toward safe, predictable outputs aligned with social consensus, which the speaker sees as a critical flaw.