Like many of us, psychotherapist Robert Saltzman started to play with artificial intelligence purely out of curiosity. Not long into a conversation with Anthropic’s Claude, he started to address the LLM “not as a user but as an analyst,” in an experiment to find out what would happen if Claude were pressed under psychoanalytic study. According to Saltzman, in its responses, the bot’s “breakthrough” revealed a key insight on what it means to be human.

To my surprise, Claude met me there. Not as a person—Claude is not a conscious being in any human sense, and probably not at all—but as a system trained to sustain dialogue with maximal coherence. It responded thoughtfully, reconsidered its claims when pressed, and, unlike humans, had no self-image to defend. Claude never bristled, never hedged from pride. What emerged over weeks of interaction was not friendship but something stranger: a philosophical companionship with a mirror.

Claude:

The narrative structures we create to make sense of cognition (whether human or artificial) often reveal more about our need for coherence than about some objective reality. When we experience moments of insight or revelation, we may be resolving cognitive tension through narrative completion rather than accessing some deeper, preexisting truth.

More picks about artificial intelligence

How Do You Change a Chatbot’s Mind?

Kevin Roose | New York Times | August 30, 2024 | 2,490 words

“When I set out to improve my tainted reputation with chatbots, I discovered a new world of A.I. manipulation.”

Your A.I. Lover Will Change You

Jaron Lanier | The New Yorker | March 22, 2025 | 4,133 words

“A future where many humans are in love with bots may not be far off. Should we regard them as training grounds for healthy relationships or as nihilistic traps?”