Scott Aaronson: Computational Complexity and Consciousness | Lex Fridman Podcast #130

On Simulation and Computability

• Scott Aronson states that indications strongly suggest the universe satisfies the Church-Turing thesis (is computable), though this does not prove we live in a simulation.
• A "perfect" simulation would be indistinguishable from reality, making direct evidence of an external simulator impossible to obtain.
• Aronson discusses a potential "exploit" in the universe, citing Peter Shor's joke that the incompatibility of General Relativity and Quantum Mechanics might be a deliberate "unspecified" region that could cause an "overflow error" or "blue screen of death" if probed.
• He dismisses the likelihood of the universe crashing due to human inquiry, noting it has held up for 14 billion years without such an event.

On Theories of Everything (ToE)

• Aronson distinguishes between a literal "Theory of Everything" and a fundamental theory of physics covering all fundamental interactions.
• He characterizes Eric Weinstein and Stephen Wolfram's work as ideas or hopes that could lead to a ToE, rather than completed theories.
• While the pursuit of a fundamental unification is a worthy scientific program, Aronson argues it would not explain emergent behaviors (e.g., cats, baseball) or fully resolve the "hard problem" of consciousness.
• He notes a cultural taboo in the physics community similar to AI: discussing grand, far-out dreams (like AGI or a ToE) can attract skepticism if not grounded in current mainstream frameworks.

On Consciousness and Integrated Information Theory (IIT)

• Aronson proposes the "Pretty Hard Problem" of consciousness: determining which physical systems are conscious and quantifying degrees of consciousness, distinct from the "Hard Problem" of why consciousness exists at all.
• He critiques Giulio Tononi's Integrated Information Theory (IIT), specifically the measure $\Phi$.
• Aronson argues IIT fails because it predicts that a simple, large grid of logic gates (implementing an error-correcting code) would be more conscious than a human, a conclusion he deems "obvious nonsense."
• He notes that Tononi's response to this critique was to "bite the bullet," accepting the theory's prediction that the logic gate grid is conscious rather than modifying the theory.
• Aronson emphasizes that a valid definition of a concept must match our pre-theoretic intuitions about that concept; IIT fails this test.

On Penrose and Quantum Gravity in Consciousness

• Roger Penrose's hypothesis requires three specific, highly speculative conditions:
  • Quantum mechanics is insufficient; the brain must exploit uncomputable laws of quantum gravity.
  • Quantum mechanics is not fundamental; it requires an "objective reduction" mechanism (spontaneous collapse of quantum states) at macroscopic scales.
  • This collapse is not random but influenced by consciousness to determine outcomes (free will).
• Aronson rejects Penrose's use of Gödel's Incompleteness Theorem as the foundation for this argument, citing Alan Turing's counter-argument that humans do not have an absolute guarantee of correctness when adding new axioms, just as formal systems do not.
• Most physicists argue that quantum gravity effects are screened off by quantum field theory in ordinary biological conditions, making Penrose's "antennae" requirement highly improbable.

On AI and the Turing Test

• Aronson critiques the Eugene Guzman chatbot (which claimed to pass the Turing Test) as a negligible advance over the 1960s Eliza bot, noting its inability to answer basic common-sense questions (e.g., "Is Mount Everest bigger than a shoebox?").
• He praises GPT-3 as a major advance over previous models, capable of generating passable essays and poetry by predicting the next word based on training data from the entire internet.
• However, Aronson argues GPT-3 does not yet pass a rigorous Turing Test because it lacks true reasoning capabilities and struggles with basic logic and arithmetic.
• He posits that human cognition likely combines predictive processing with other distinct mechanisms (e.g., visual reasoning) that current models lack, suggesting scaling alone may eventually hit diminishing returns.
• Aronson defines consciousness in the context of the Turing Test as relative to the observer's ability to understand the system's internal machinery; if the "black box" is transparent (e.g., a lookup table), it ceases to appear conscious to an observer who understands the mechanism.

On Computational Complexity (The "Complexity Zoo")

• Aronson created the "Complexity Zoo" website to catalog complexity classes, describing them as the "elements" of theoretical computer science.
• P: Problems solvable by a deterministic computer in polynomial time (e.g., addition, sorting).
• NP: Problems where a solution is easy to verify in polynomial time but potentially hard to find (e.g., factoring).
• P vs. NP: Aronson bets with 97-98% confidence that P $\neq$ NP, meaning not all easily verifiable problems are easily solvable.
• BQP: Problems efficiently solvable by quantum computers; Aronson places BQP within PSPACE, noting quantum computers cannot solve uncomputable problems (like the Halting Problem) but can solve certain NP-like problems faster.
• SZK (Statistical Zero Knowledge): Aronson cites this as one of the most beautiful classes, encompassing problems that can be proven true without revealing why they are true (zero-knowledge proofs).

On the Pandemic and Societal Institutions

• Aronson describes the institutional failure (CDC, media) during the pandemic as a "historic failure" and a "staggering" lack of imagination, noting that even the movie Contagion failed to predict the level of incompetence and conspiracy theory influence on leadership.
• He argues a "World War II mentality" of unified action and rule-breaking for the greater good was missing, exacerbated by political division.
• He expresses deep fear regarding the potential for violence and instability following the election if results are contested.
• Aronson is a vocal opponent of "cancel culture," arguing it is ineffective at combating racism/sexism and destroys open discourse by enforcing rigid orthodoxy.
• He advocates for a "positive" approach, urging people to engage in open discourse and express support for reasonable dialogue without fear of social retribution.

On Love and Meaning

• Aronson confirms that love for family and life has played a central, grounding role in his existence.
• He acknowledges that as a theoretical computer scientist, he cannot offer deep technical insights into the nature of love, humorously noting that standard mathematical tools (Fourier transforms, integrals) are useless for analyzing the heart.