#317 — What Do We Know About Our Minds?

Episode Summary

Executive Summary: Sam Harris opens with a warning that GPT-4 has accelerated his concerns about AGI, alignment, and the flood of AI-generated misinformation, then speaks with psychologist Paul Bloom about his new book Psych and broader questions about the mind. They discuss psychology’s strengths and limits, fiction as a superior guide to human experience, consciousness vs. intelligence, social media’s distortions, truth vs. belonging, authority in science, and why our increasingly mediated information environment may destabilize public reason.

Main Topics: AI, AGI, and Alignment Risk (Priority: 5/5): Harris argues GPT-4 has made AI risks more immediate: powerful systems are being deployed before society understands their implications, worsening both existential alignment concerns and near-term misuse via hoaxes and misinformation. Social Media, Misinformation, and Attention Collapse (Priority: 5/5): Both speakers describe Twitter and algorithmic feeds as psychologically corrosive, amplifying outrage, distorting reputations, and fragmenting attention; Harris frames social media and AI as a combined threat to shared reality. Psychology, Fiction, and Understanding the Mind (Priority: 4/5): Bloom presents Psych as an accessible survey of psychology, while both agree that novels, films, and TV often reveal human motives and relationships better than textbooks or studies. Truth, Bullshit, and Social Belonging (Priority: 5/5): The conversation distinguishes lying from bullshitting and explores why people often conform to community beliefs for social survival, even when those beliefs conflict with truth. Consciousness, Intelligence, and Machine Minds (Priority: 4/5): They separate intelligence from consciousness, noting that AI may be genuinely intelligent without being conscious, while the moral status of potentially sentient systems remains unresolved. Science, Authority, and the Limits of Reductionism (Priority: 4/5): They discuss how science relies on authority as a practical shortcut, yet remains vulnerable to error; they also reject crude reductionism that would replace psychological explanations with only neural terms. Well-being, Happiness, and Real-World Life (Priority: 3/5): Harris describes a renewed desire for in-person life, nature, and audio-based work; Bloom notes that online communities can also provide real social connection, even as the medium can become addictive or harmful.

Key Arguments: GPT-4 and rapid AI progress have not reassured Harris about alignment; instead they suggest society is racing ahead without adequate safeguards. The biggest near-term AI danger may be an epistemic collapse: floods of fake text, audio, images, and video could make the internet less trustworthy. Consciousness and intelligence are distinct problems; AI can be intelligent without being conscious, and the ethical stakes differ depending on whether machines can suffer. Twitter and similar platforms are psychologically distorting because they magnify outrage, create false impressions of other people, and reward compulsive attention. Fiction often conveys the texture of human life better than psychology research because it captures lived complexity, motives, and relationships. Bullshitting is more dangerous than lying in public life because the bullshitter is not constrained by truth-tracking at all. People often adopt false beliefs for social belonging; in many communities, conformity can be rational even when beliefs are objectively wrong. Science depends on authority and consensus as practical heuristics, but those must remain revisable and should not become dogma. Reductionist accounts that replace mental terms with neural terms miss the levels at which human life is actually understood and lived. AI may eventually become a trusted gatekeeper or filter for misinformation, but current models already hallucinate and cannot be relied on as fully trustworthy sources.

Data Points: Year of Harris's TED Talk on AGI: 2016 - Harris references his earlier TED Talk as an anchor for his longstanding concerns about AGI. Podcast subscriptions required for full episodes: 100% of full episodes behind subscriber feed - The episode begins and ends with a subscriber-only notice. Books authored by Paul Bloom: 8 - Harris introduces Bloom as the author of eight books, including Psych. Major award mentioned: $1 million - Bloom is noted as the recipient of the Klaus J. Jacobs Research Prize. Twitter tenure: 12 years - Harris says he was on Twitter for 12 years before leaving. Estimated share of bad life events linked to Twitter: ~100% - Harris says 'something like a hundred percent' of the truly bad things in his life over the last 10 years came from Twitter. Estimated alternative share of bad life events linked to Twitter: 90% still likely too low - He says even 90% would underestimate Twitter's impact. Probability of AI killing us all: 5%–10% - Harris references his earlier claim that even low single-digit existential risk is worth serious attention.

Pivotal Quotes: "With GPT-4, we've blown past all of that." — Sam Harris: Harris argues that AI deployment has already exceeded the caution he and others expected on the way to human-level intelligence. "The difference between a liar and a bullshitter is that a liar has to be tracking what the truth is... whereas the bullshitter is just talking." — Sam Harris: Harris explains why bullshitting is particularly corrosive in politics and public discourse. "The mind is the brain." — Paul Bloom: Bloom summarizes a core scientific commitment while emphasizing that this does not reduce away higher-level psychological explanation.

Implications: The episode frames AI and social media as twin forces eroding trust, attention, and shared reality. For listeners, the takeaway is to value verified sources, protect attention, and expect psychology—not just tech—to shape the next phase of public life.