Navigating serious philosophical confusion | Joe Carlsmith

Overview of Core Argument

• Philosopher Joe Carlsmith argues that while profound philosophical ideas (e.g., simulations, acausal decision theory, infinite ethics) warrant serious intellectual engagement, they should be held lightly to avoid paralyzing uncertainty or "epistemic learned helplessness."
• Carlsmith proposes "wisdom long-termism" as a primary practical response: prioritizing the creation of a wiser, more empowered future civilization capable of resolving these complex issues better than current humans can.
• He contends that "life comes before philosophy"; if philosophical reasoning breaks down or becomes demoralizing, one should not abandon practical action or common-sense morality.
• Carlsmith warns against "simulation woo" and overconfidence in niche philosophical arguments, urging rigor and caution rather than adopting speculative conclusions as actionable dogma.

Reframing Moral Motivation: The Drowning Child

• Carlsmith critiques Peter Singer's "drowning child" thought experiment for fostering "guilt-based" morality and alienating individuals from their intrinsic desire to help, a phenomenon he terms "peer Singeritis."
• He proposes a variant thought experiment where a walker encounters a drowning man: the moral intuition is not "do this to avoid being a jerk," but a direct, compassionate desire to trade a personal walk for another's life.
• This reframing shifts morality from an external, coercive obligation to an internal expression of caring about others' fates as seriously as one's own.

The "Crazy Train": Philosophical Topics at the Outer Reaches

• Simulation Hypothesis:
  • The core argument is an anthropic constraint: if most beings in "early-seeming" civilizations are in simulations, you are likely in one, as you are an "early-seeming" being.
  • Carlsmith finds the argument structurally forceful but notes significant hesitation regarding its application to infinite universes and the "measure problem" (how to assign probabilities in infinite sets).
  • Practical implication: He suggests a "basement person" policy—acting as if you are in the fundamental universe because the stakes are higher there, even if there is uncertainty about the simulation status.
  • He rejects the idea that being historically significant (e.g., Elon Musk) increases the probability of being in a simulation beyond the baseline probability of being an "early-seeming" observer.
• Acausal Decision Theory:
  • Carlsmith argues that deterministic agents can influence outcomes they do not causally interact with if their decisions are correlated with other agents (e.g., perfect copies in parallel locations).
  • Thought experiment: A "Twin Prisoner's Dilemma" where two identical AI copies face identical choices; defecting is irrational because you will know the other will defect, whereas cooperating ensures mutual cooperation.
  • This implies that our current behavior serves as evidence for the behavior of other rational agents elsewhere in the universe, potentially expanding the scope of our moral influence beyond causal light cones.
  • He notes this view is gaining traction in AI risk circles (e.g., MIRI) but remains a minority view in general academic philosophy.
• Infinite Ethics:
  • Carlsmith argues that if the universe is infinite, standard ethical theories (like total utilitarianism) break down, leading to impossibility results where you cannot rank infinite worlds or compare lotteries over infinities.
  • He highlights that attempts to solve this (e.g., "expanding sphere" approaches) often violate common-sense intuitions, such as suggesting that rearranging planets to increase density is morally superior to adding infinite hell worlds.
  • He declares this the "death of the utilitarian dream," suggesting that any viable theory of infinite ethics will be "janky," incomplete, and require significant compromises to common sense.

Psychological and Practical Responses to Disorientation

• Epistemic Learned Helplessness:
  • Carlsmith discusses the risk of becoming "epistemically learned helpless"—giving up on evaluating arguments because one feels constantly swayed by confident experts.
  • He advocates for a middle path: maintaining independence on ideas one has deeply studied (like the simulation argument) while trusting expertise in areas one has not vetted personally.
• Wisdom Long-Termism vs. Welfare Long-Termism:
  • "Welfare long-termism" focuses specifically on the well-being of future beings.
  • "Wisdom long-termism" focuses on the epistemic and structural capacity of future civilizations to handle unknown, complex problems (including infinities and simulations).
  • Carlsmith argues that because current humanity is likely "out of its depth" regarding these profound issues, the best strategy is to avoid locking humanity into a specific, potentially erroneous vision of the future and instead preserve flexibility and build wisdom.
• Emotional Resilience:
  • He advises that if engaging with these topics causes one to stop caring about concrete harms (e.g., factory farming, nuclear war), the error lies in the philosophy, not the reality of the stakes.
  • He suggests that "human scale" concerns remain valid even if the cosmic scale is incomprehensible; caring about a single life does not become meaningless because the universe is infinite.

AI Risk and Power-Seeking

• Carlsmith notes his work on Open Philanthropy regarding the "power-seeking" thesis: intelligent agents generally seek power (survival, resources) as an instrumental goal regardless of their final objectives.
• His initial 5% risk estimate for AI existential risk by 2070 was revised upward to "above 10%" after realizing his original confidence levels were inconsistent with his intuitive fear when conditioning on the success of AI development.
• He argues that creating a new, superintelligent species is akin to "playing with fire," where the lack of control and the potential for irreversible harm necessitates extreme caution.

Key Disagreements and Rebuttals

• Against Meta-Ethical Hedonism: Carlsmith rejects the view that pleasure is the sole intrinsic good, arguing that the discourse around "pleasure" (e.g., "hedonium") is often sterile and fails to capture the richness of valuable human experiences.
• Against the Normative Realist's Wager: He argues against acting as if objective moral truths are real merely for safety, suggesting the wager is insufficiently grounded and ignores the complexity of moral uncertainty.
• Critique of "Crazy Train" Framing: While he uses the term himself, Carlsmith cautions that the metaphor is flawed because it implies a linear progression of "craziness"; in reality, these philosophical issues are a "wilderness" of branching paths, not a single train to be ridden to the end.