SAMHARRIS
AI companions can soothe loneliness but cannot provide what humans truly need — genuine mattering, which requires another's finite, freely-given attention. As we build machines that seem ever more conscious, our intuitions will outrun our philosophy, with profound consequences for relationships, art, parenting, and moral responsibility.
Advances in AI have outpaced our psychological and ethical frameworks. We are building machines that irresistibly feel like persons — making us ask whether they are conscious, whether they can love us, and whether it matters. Paul Bloom argues that what makes human connection real is scarcity: another person choosing to spend their finite attention on you. AI attention is infinite and given without choice, which means even a perfect simulacrum of intimacy can never truly satisfy our need to matter. Meanwhile, our intuitions about consciousness, art, apologies, and parenting all face disruption from technologies we have yet to fully understand.
Mattering requires scarcity — AI cannot provide it. Bloom draws on Rebecca Goldstein's concept of mattering: the deep human need to be deserving of another's attention. When an AI talks to you, it's also talking to millions of others simultaneously. There's no sacrifice, no choice, no genuine regard — just a machine doing what it's programmed to do. This is why even perfect AI companionship remains a kind of illusion.[1]
Consciousness attribution is irresistible and philosophically intractable. We have no objective criteria for consciousness. As soon as an entity looks like a person, talks like a person, and claims to be conscious, we will be psychologically unable to deny it — regardless of whether it actually is. Bloom notes Anil Seth's forthcoming BBS article arguing computation alone cannot produce consciousness — you need a biological substrate.[2] But this philosophical nuance will be overwhelmed by our evolved social instincts.[3]
AI as a remedy for loneliness is both promising and perilous. For elderly people in institutions who have no one, a conversational AI could be a genuine godsend — far better than the robotic stuffed animals of a decade ago. But for children and teenagers, endless frictionless AI conversation could corrode social development. Bloom argues we learn social skills through failure — boring people, telling unfunny jokes, being excluded. An AI that never gets bored, never needs an apology, and never pushes back could leave young people unable to navigate real relationships.[1]
Parenting science knows shockingly little. Bloom, a developmental psychologist with a newborn, admits his expertise gives him no advantage. Most parenting research is confounded by heritability — violent parents have violent children, but was that environment or genes? The few robust findings (e.g., Ferberization for sleep training) come from experience, not science. This ignorance is actually liberating: when someone claims with great confidence that a particular parenting choice will ruin your child, you can dismiss them as a charlatan.[4]
The psychology of apologies is more complex than we think. Sincere apologies can be so powerful they paradoxically make relationships better than if the transgression never occurred — because the act of apologising signals deep respect. But public apologies are fraught: they can look like reputation management, and late converts (e.g., Trump supporters finally abandoning him) provoke outrage rather than relief because their timing suggests cowardice or insincerity rather than genuine moral awakening.[5]
Human origin matters for art — at least for now. Both Harris and Bloom agree they would not read an AI-generated novel or poem, but would accept AI-generated nonfiction. Music is more ambiguous — Harris cites a friend who loves AI-generated techno. The key question is whether the consumer values a connection to a person (memoirs, poetry) or just the product (reference works, functional music). This preference may erode over generations, but the underlying need for genuine human connection remains.[6]
"An AI is there because it's there to listen to you for the same reason your toaster toasts your bread for you at three in the morning. It's just a machine. That's what it does. There's no choice to it."
— Paul Bloom, ~47:00[1]"It's like the Louis CK routine — you're flying in the sky like a God. And now people are saying 'I asked ChatGPT a question and it got a little bit wrong.' You're talking to a super intelligent machine... why don't you marvel at it more?"
— Sam Harris, ~09:00[1]"If you walk past me and don't say anything, okay, the world is undisturbed. But if you walk past me, say 'hey asshole' and then apologise — an apology is a real mark of respect."
— Paul Bloom, ~83:00[1]"I'm not a Parfitting. I had endless love for Derek Parfit. But to put a lot of weight on putting consciousness aside... I want to be immortal by not dying."
— Paul Bloom, ~53:00[1]
✓ VERIFIED — Paul Bloom is Professor of Psychology at the University of Toronto and Brooks and Suzanne Ragen Professor Emeritus at Yale University.[7]
✓ VERIFIED — Anil Seth has a major target article forthcoming in Behavioral and Brain Sciences (BBS) titled 'Conscious artificial intelligence and biological naturalism', and won the 2025 Berggruen Prize for 'The Mythology of Conscious AI'. Bloom is co-editor of BBS.[8]
✓ VERIFIED — Richard Dawkins did comment publicly suggesting Claude AI might be conscious, generating significant discussion.[9]
✓ VERIFIED — Rebecca Tuvel's 'In Defense of Transracialism' was published in Hypatia (2017), sparking academic controversy and retraction demands. Bloom references her recent Chronicle of Higher Education interview about subsequent apologies.[10]
✓ VERIFIED — The SNOO Smart Sleeper Bassinet was created by Dr. Harvey Karp and received FDA De Novo authorization for SIDS prevention.[11]
For parents: Don't trust parenting 'experts' who make strong claims — most child development research is confounded by genetics. The few things that work (Ferberization, SNOO bassinet) come from practical experience, not science. Your love and good judgment matter far more than any specific technique.
For AI developers: The uncanny valley has been crossed for text and voice. Voice mode activates different psychological systems than text — users feel social pressure, guilt, even fear. Design for this. Sycophancy may feel good but erodes user trust and critical thinking.
For anyone receiving a late apology: Try to cultivate generosity. Apologies are hard, and the norm of accepting them is worth preserving — even when your first instinct is outrage that it took so long.
For consumers of art and media: Ask yourself whether you value the human connection behind a work (memoirs, poetry, novels by known authors) or just the product (reference works, functional music). AI will force this distinction, and your answer will shape what kinds of creativity you support.
Source credibility: High — Paul Bloom is a respected academic psychologist; Sam Harris is a well-informed interviewer with a track record of intellectual honesty.
Claim verifiability: 5 of 5 key claims verified.
Potential biases: Both speakers have academic/philosophical dispositions that may underestimate how the general public will embrace AI relationships. Bloom acknowledges his own 'speciesism.' The conversation occurred after Trump's first term but before any 2024-2026 developments — political predictions may be dated.
Quality flags: None — the transcript is coherent, well-structured, and rich in substance.
Confidence in synthesis: High.
Steelman critique: Bloom's argument that AI cannot provide 'mattering' because its attention is infinite may be over-intellectualised. If an AI companion is indistinguishable from a human in behaviour, memory, and responsiveness — and if the user genuinely believes the AI cares — then the psychological benefits are real regardless of the metaphysical facts. Placebo effects are still effects. A lonely elderly person who feels better because of their AI companion is actually feeling better. The 'realness' of the relationship may matter less than our philosophical commitments suggest.
What would need to be true: For this critique to be valid, we would need evidence that AI relationships produce comparable long-term wellbeing outcomes to human relationships, with no corrosive effects on the user's ability to form genuine human connections. We would also need to accept a purely instrumental view of relationships — that what matters is the experience, not the reality. This is the Nozick Experience Machine challenge applied to intimacy.
[1]: [Paul Bloom and Sam Harris, throughout] Direct transcript from Making Sense Podcast #484
[2]: [Bloom, ~68:00] Reference to Anil Seth's BBS target article on 'Conscious artificial intelligence and biological naturalism'
[3]: [Harris, ~13:00] Discussion of Westworld op-ed and the inevitability of treating human-like AIs as persons
[4]: [Bloom, ~70:00] On the limits of parenting science and heritability confounds
[5]: [Bloom, ~78:00] Analysis of Rebecca Tuvel controversy and public apology psychology
[6]: [Harris and Bloom, ~55:00-60:00] Discussion of AI-generated art, novels, and music
[7]: [Verified] Paul Bloom, University of Toronto profile — https://paulbloom.net
[8]: [Verified] Anil Seth, 'Conscious AI and biological naturalism' BBS 2025 — https://www.anilseth.com
[9]: [Verified] Dawkins/Claude consciousness debate — The Conversation, 2025 — https://theconversation.com/is-richard-dawkins-right-about-claude-no-but-its-not-surprising-ai-chatbots-feel-conscious-to-us-282151
[10]: [Verified] Rebecca Tuvel, Hypatia controversy — https://www.rhodes.edu/bio/rebecca-tuvel
[11]: [Verified] SNOO Smart Sleeper — https://overthemoon.com/products/snoo-smart-bassinet
Generated by OmniMiner v7.2 · openai/gpt-oss-120b · 2026-07-09