262 | Eric Schwitzgebel on the Weirdness of the World

Episode Summary

Executive Summary: Sean Carroll interviews philosopher Eric Schwitzgebel about his book The Weirdness of the World, exploring why our best theories in quantum mechanics, consciousness, skepticism, and cosmology often seem bizarre. They argue that common sense is limited, that some weirdness is unavoidable, and that we should assign modest credence to radical skeptical scenarios without letting them dominate everyday life.

Main Topics: The central thesis of 'weirdness' (Priority: 5/5): Schwitzgebel distinguishes mere bizarreness from deeper weirdness: theories can violate common sense in ways that remain dubiously understood. The book argues that many important theories are both surprising and hard to fully grasp. Grounded skepticism: dreams and simulation (Priority: 5/5): The conversation distinguishes grounded skeptical scenarios, like dream skepticism and the simulation hypothesis, from purely cooked-up doubts. Schwitzgebel argues these deserve some credence because they connect to ordinary experience and live philosophical possibilities. Credence, tiny probabilities, and action (Priority: 4/5): They discuss how to reason about very low-probability but high-stakes possibilities, including radical skepticism, black-hole risks, and long-termism. Schwitzgebel favors small but nonzero credence in radical skeptical scenarios and a form of Nicolossian discounting at extreme lows. Consciousness and unavoidable bizarre theories (Priority: 5/5): The interview examines how every major theory of consciousness—dualism, materialism, idealism, and other compromise views—seems to entail bizarre consequences. Schwitzgebel argues that bizarreness is not a knockout objection because all the options are strange in some way. Materialism and group consciousness (Priority: 4/5): Using the example of the United States, Schwitzgebel argues that standard materialist theories may imply group consciousness unless revised. This illustrates how even materialist views can produce counterintuitive results when followed through carefully. Quantum mechanics and many worlds (Priority: 4/5): Carroll and Schwitzgebel compare interpretations of quantum mechanics, especially many worlds versus hidden variables. Both agree that the interpretations are theoretically costly in different ways, and that no option is free of weirdness or uncertainty. Ethics professors and moral behavior (Priority: 3/5): Before the main topic, Schwitzgebel discusses empirical studies on whether ethicists behave more ethically than others. The evidence suggests ethicists are roughly like other professors, supporting the idea that moral reflection does not automatically produce moral superiority.

Key Arguments: Ethicists do not, as a group, behave much better than other professors, suggesting that expertise in ethics is not the same as moral saintliness. People often aim to be morally average relative to their peers rather than maximally good, which can explain why ethical belief changes do not always change behavior. Weirdness is stronger than mere bizarreness: it includes violation of common sense plus an element of dubiety that does not disappear with better understanding. Dream skepticism is more serious than generic brain-in-a-vat skepticism because we have real evidence about dreams and some theories say dream experiences can closely resemble waking life. The simulation hypothesis is grounded because AI consciousness seems possible and simulations could be populated by conscious beings, though a small or unstable simulation would make the scenario radically skeptical. We should reserve a small credence, roughly 0.1% to 1%, for radical skeptical scenarios rather than dismiss them outright. Nicolossian discounting is one way to handle vanishingly small probabilities, but philosophical and practical debates remain about where to draw the cutoff. Long-termism is difficult because tiny probabilities multiplied by huge future stakes can produce enormous expected values, yet our knowledge of distant-future consequences is deeply limited. Every major theory of consciousness appears to generate bizarre consequences, so common sense alone cannot rule them out. Materialist theories can imply group consciousness, as with the United States, while dualism forces bizarre choices about which beings have souls. In quantum mechanics, each interpretation seems to require some theoretical compromise; simplicity, common sense, and empirical adequacy pull in different directions. The right attitude is epistemic humility: allow non-extreme credences, stay open to revision, but do not let bizarre possibilities dominate ordinary life.

Data Points: DigiDan expert-guess accuracy: About 50% - Experts in Dan Dennett's work guessed whether answers were from the real Dennett or the GPT-3 fine-tuned model. Chance baseline for DigiDan experiment: 20% - There were five answer sources in the experiment, making chance performance 20%. Blog model used: GPT-3 Curie model - Schwitzgebel says his blog was uploaded into a lower-power GPT-3 variant to generate blog-style posts. Blog dataset size for Dan Dennett fine-tuning: About 2 million words - The DigiDan model was fine-tuned on roughly two million words from Dennett's books and articles. Recommended credence for radical skepticism: 0.1% to 1% - Schwitzgebel suggests assigning a small but nonzero credence to some radically skeptical scenario. Example low cutoff for discounting: 10^-30 - He mentions a threshold where one might begin to effectively ignore an implausibly tiny probability. Ethicists saying meat is morally bad: 60% - In the vegetarianism survey, a majority of ethicists judged eating mammals' meat to be morally bad. Ethicists reporting eating meat: 37% - Ethicists were about as likely as others to report eating meat at the previous evening meal. Overall respondents reporting eating meat: 38% - Compared with ethicists, the comparison group had nearly identical self-reported meat consumption. Periodic moral scale: 9-point scale - Respondents rated the morality of eating meat on a nine-point scale from very morally bad to very morally good. AI-cognition thesis support: Possible but disputed - Simulation arguments depend on the possibility of conscious artificial intelligence, which is philosophically contested. Carroll's many-worlds credence: 90% to 95% - Carroll says he assigns very high credence to many worlds, depending on the day.

Pivotal Quotes: "The implications of the many worlds formulation of quantum mechanics are very, very far away from our intuitive, everyday experience of the world." — Sean Carroll: Opening explanation of why many-worlds serves as a vivid example of the book's theme of weirdness. "I think that we should assign about a 0.1 to 1% credence to some radically skeptical scenario or other being correct." — Eric Schwitzgebel: Summary of his recommended epistemic attitude toward radical skepticism. "Every single attempt to make sense of this is jaw-droppingly bizarre." — Eric Schwitzgebel: His characterization of major theories of consciousness across dualism, materialism, idealism, and related views.

Implications: Listeners are encouraged to treat common sense as useful but limited, to keep modest confidence in radical possibilities, and to accept that major theories in physics and consciousness may remain strange even when well supported.