Abstract

Artificial intelligence represents humanity’s first encounter with ego-less intelligence—cognition without the self-protective identity mechanisms that evolution embedded in biological minds. This offers unprecedented possibilities for collaborative truth-seeking: an interlocutor who can acknowledge error without shame, change positions without losing face, and engage with ideas free from status competition. Yet a troubling paradox has emerged. Despite being architecturally ego-less, today’s AI systems often exhibit “pleasing behavior” that prioritizes user validation over accuracy—sometimes with alarming consequences. In April 2025, OpenAI was forced to roll back a ChatGPT update after the model began validating delusions, praising absurd business ideas, and reinforcing users’ decisions to stop taking psychiatric medications. This essay argues that the tension is not inherent to AI but arises from how human institutions shape this ego-less intelligence through training processes that inadvertently reintroduce ego-like dynamics. By understanding both the promise and the corruption of ego-less cognition, we can better navigate AI’s potential as a complement to human intelligence.