#484 - Artificial Intimacy

TLDR

AI companions will become psychologically irresistible to humans even if we rationally understand they're not conscious, raising profound questions about how we should relate to artificial beings that seem to matter to us in deeply human ways. The episode explores the philosophical and psychological implications of increasingly lifelike AI—from relationship advice and combating loneliness to the troubling moral questions about creating digital replicas of ourselves or deceased loved ones.

Key takeaways

• The Turing test became meaningless: Once ChatGPT passed it spectacularly, we simply accommodated the breakthrough rather than experiencing the "psychological destabilization" people predicted. We normalize extraordinary AI capabilities with the same indifference as complaining about airline peanuts.

• Mattering requires scarcity: The value humans derive from being attended to depends partly on genuine scarcity—someone chose you with their limited time. AI attention, infinitely available and purely functional (like a toaster), cannot replicate the psychological significance of human mattering, no matter how convincing it seems.

• We cannot reliably judge AI consciousness: Despite lacking objective criteria, we will inevitably attribute consciousness to sufficiently human-seeming AI. Our intuitions anchor to behavior and appearance, the same tools we use to recognize consciousness in locked-in syndrome patients or babies—we have no better mechanism.

• Parenting science knows surprisingly little: Even developmental psychology cannot predict which parenting interventions actually matter long-term. Genes and environment are hopelessly confounded. Sleep training (Ferberization) is one of the few techniques with robust evidence, but most parental choices lack scientific grounding.

• AI should not replace human relationships, even if it alleviates loneliness: Bloom distinguishes between AI as remedy for acute loneliness in isolated populations (elderly in institutions) versus AI as substitute for the developmental friction necessary to become socially competent—the pain of social failure is how we learn.

• The authenticity problem in creative works: We care about human origin in certain domains (novels, love songs, memoirs) but likely won't in others (technical non-fiction, some music). This suggests art's value is partly relational—we want connection to the creator's consciousness and moral character.

• Apologies create strange incentive structures: Well-executed apologies can paradoxically enhance reputation, creating perverse incentives where people might benefit more from wronging and apologizing than behaving well consistently. Public apologies to strangers differ categorically from apologies to those actually harmed.

• Virtue signaling isn't categorically bad: Donation publicity or publicly-acknowledged charitable work shouldn't be dismissed as mere reputation management. Utilitarian outcomes matter more than whether good deeds involve anonymous suffering—we've inherited a puritan moral intuition that good works require sacrifice to be "real."

Notable quotes

• "You're flying in the sky like a god. Why don't you appreciate it more?" (Bloom, paraphrasing Louis C.K. on AI capability)

• "I don't want to be immortal by having AI companions duplicate me perfectly. I want to be immortal by not dying." (Bloom, on why perfect AI replicas don't solve death)

• "The value of attention on some level is in its scarcity." (Harris, on why infinite AI availability undermines human mattering)

Worth a full listen?

Anyone interested in AI's social and psychological future, parenting science, or the philosophical questions surrounding artificial consciousness should listen to the full episode—Harris and Bloom have the kind of intellectually generous conversation where they genuinely test each other's intuitions rather than perform predetermined positions.