Dwarkesh PodcastScott Aaronson - Quantum Computing, Complexity, and Creativity
CHAPTERS
- 0:00 – 0:43
World-expert mindset: master a tiny problem
Aaronson opens with a pragmatic philosophy of expertise: becoming broadly expert can take decades, but becoming the world expert on a narrowly defined problem can be surprisingly fast. He frames this as a powerful strategy for learning, publishing, and building momentum.
- •Broad expertise is slow; narrow expertise can be achieved quickly
- •Pick a small, well-scoped question and push it further than anyone else
- •Turning narrow mastery into writing/projects creates compounding opportunities
- 0:43 – 4:51
Leaving high school early: GED, Hong Kong, and escaping a bad fit
Aaronson corrects the “graduated at 15” narrative: he earned a GED after a rocky high-school experience. A year in Hong Kong, curriculum mismatch, and limited math options pushed him toward skipping grades and seeking an environment better aligned with his interests.
- •Social and academic unhappiness in high school as the driving force
- •Hong Kong year and curriculum mismatch leads to grade-skipping
- •Running out of math classes and the school rejecting alternatives (EPGY)
- •Decision to leave for an early college-style program
- 4:51 – 7:20
Clarkson School and the bumpy path to Cornell
He describes attending the Clarkson School (a one-year bridge to college), starting research early, and then facing widespread college rejections due to an unconventional background. Cornell acceptance came with a bureaucratic hurdle: no high school diploma, requiring an exception to obtain a GED at 15.
- •Clarkson as a ‘college courses while in high school’ escape hatch
- •Early exposure to professors and research
- •Unusual background leads to many rejections; Cornell/CMU accept
- •Diploma/GED requirements and the phys-ed bureaucratic snag
- 7:20 – 9:21
Accelerating academically vs. paying socially
Aaronson explains that most classmates barely noticed his age, so academics weren’t the main problem. The real cost was social development—dating, driving, and general life skills lagged—creating stress despite academic success.
- •Age difference mattered less in class than in social life
- •Parents warned about social costs; warnings proved accurate
- •‘Weird order’ of life milestones (PhD before driving/social life)
- •Tradeoff: social misery already existed, so he optimized for learning
- 9:21 – 19:21
Specializing early and the mystery of ‘miracle years’
He reflects on whether learning fundamentals early confers special cognitive advantages beyond having more time. The conversation broadens to why young scientists sometimes have bursts of major discoveries, and whether aging effects are cognitive, motivational, or time-constraint driven.
- •Early specialization: freedom to study what matters vs. rigid schooling
- •Teenhood as a modern construct; historically teens apprenticed
- •Miracle years (Newton/Einstein) and ‘ideas in the air’ dynamics
- •Older researchers: disentangling brain slowdown from time/motivation
- 19:21 – 26:40
Why quantum information ideas arrived so late (1920s → 1990s)
Aaronson answers why quantum teleportation and quantum computing were discovered decades after quantum mechanics. He traces missing prerequisites: viewing entanglement as a usable resource, the birth of classical computing and complexity theory, wartime disruptions, and shifting fashions in physics.
- •Bell’s 1960s work reframed entanglement as testable/usable
- •Wiesner’s quantum money/crypto ideas were ahead of publishability
- •1920s–30s focus: applying QM to chemistry; CS foundations still forming
- •WWII and postwar particle/QFT focus pushed foundations out of fashion
- •Complexity theory (P/NP) emerged in 1960s–70s as a prerequisite
- 26:40 – 30:26
Deutsch, many-worlds, and alternative routes to quantum computing
Dwarkesh raises Deutsch’s claim that taking many-worlds seriously enabled quantum computing. Aaronson grants this was Deutsch’s personal motivation, but notes Feynman and others arrived independently via simulation and physics-of-computation lines of thought, not interpretational commitments.
- •Deutsch’s many-worlds framing as a conceptual ‘shake’ to orthodoxy
- •Feynman’s motivation: efficient simulation of quantum physics
- •1970s–80s physics of computation: reversibility/thermodynamics influences
- •Quantum computing had multiple independent intellectual tributaries
- 30:26 – 38:00
Are outsiders and ‘crazy ideas’ important? Academia, arXiv, and the boundary cases
Aaronson pushes back on the idea that modern academia is closed to new ideas—at least in his areas—arguing arXiv lowered publishing barriers and increased idea volume. He then distinguishes productive outsiders (often with partial formal training) from crank dynamics, sharing extreme anecdotes and a positive example of blog-driven collaboration.
- •In some fields innovation is harder; in theoretical physics/CS arXiv opened floodgates
- •Today’s bottleneck is filtering too many bold ideas, not gatekeeping
- •Outside-academia breakthroughs often come from ‘near-academia’ backgrounds
- •Crank spectrum: misunderstandings → conspiratorial escalation → threats
- •Blog readership as a bridge; busy beaver survey sparked a hobbyist collaboration
- 38:00 – 45:37
Busy Beaver: defining the fastest-growing numbers and why only a few values are known
Aaronson gives a detailed, accessible explanation of the busy beaver function: among n-state programs that halt, take the one that runs the longest. He explains why it outgrows every computable function, connects it to the halting problem and Gödel incompleteness, and highlights the shockingly small set of known exact values.
- •Definition via halting programs/Turing machines: max steps before halting
- •Avoiding paradoxes of ‘largest named number’ using program-based naming
- •Busy beaver dominates any computable function (and quickly)
- •Connections: uncomputability, halting problem, Gödel incompleteness
- •Known values through n=4; explosive lower bounds for n=5,6,7
- 45:37 – 55:28
Busy Beaver independence and extending set theory: engineering contradictions into small machines
Dwarkesh asks whether we can keep extending axioms to prove larger busy beaver values. Aaronson describes a concrete research program: build a small Turing machine that halts iff it finds a contradiction in set theory, implying limits on what the theory can prove; he cites improvements from ~8000 states down to <800 by a hobbyist, and discusses large cardinal axioms and the inevitability of escalating complexity.
- •Independence question: smallest n where BB(n) is independent of standard set theory
- •Construction: machine enumerates proofs and halts only on contradiction
- •Aaronson/Yedidia result (~8000 states) and later hobbyist improvement (<800)
- •Large cardinal axioms as a route to stronger theories
- •No systematic search for ever-stronger consistent axioms (or BB becomes computable)
- 55:28 – 1:02:22
Quantum algorithms after Shor/Grover: motifs, low-hanging fruit, and new problem discovery
Aaronson argues Shor’s and Grover’s algorithms may be foundational design motifs akin to dynamic programming or linear programming in classical CS—discovered early because they’re fundamental. He suggests the path to new quantum speedups may require identifying new problem formulations and applications, and notes a concrete open candidate: faster quantum edit distance.
- •Classical algorithms reuse early-discovered motifs (recursion, DP, LP, elimination)
- •Shor/Grover as ‘basic motifs’ of the quantum algorithmic universe
- •Challenge: people demand new algorithms without naming target problems
- •Possibility: new algorithms emerge from new problem statements/applications
- •Candidate frontier: subquadratic quantum algorithm for edit distance
- 1:02:22 – 1:06:04
Why innovations cluster: Bell Labs, cultural hubs, and idea feedback loops
The conversation turns to why breakthroughs often emerge from tightly connected groups and places. Aaronson uses Bell Labs and historical intellectual centers (Athens, Florence, Cambridge, Silicon Valley) to contrast two explanations: environments that catalyze ideas vs. selection effects that attract unusually capable people.
- •Bell Labs as a late-stage example even for Shor/Grover-era excitement
- •Clusters as idea-amplifiers: inspiration, competition, rapid feedback
- •Alternative explanation: hubs attract the kind of people who innovate
- •Correlation vs. causation remains hard to disentangle
- 1:06:04 – 1:13:28
Complexity meets economics: Nash equilibrium hardness and limits of rationality
Aaronson explains theoretical results suggesting computing Nash equilibria is computationally hard in a precise sense (PPAD-style guarantees), even though equilibria always exist. He links this to a broader lesson for economics: existence theorems don’t imply agents or markets can feasibly compute outcomes, and distinguishes computational limits from Hayek’s information/knowledge limits.
- •Two-player zero-sum equilibria are easy (linear programming); Nash is harder
- •Nash existence via fixed-point theorems doesn’t yield efficient algorithms
- •Hardness result: complete for a class of guaranteed-solution problems
- •Implication: if equilibria require infeasible computation, markets/agents won’t find them
- •Separating lack of information from inability to compute on information
- 1:13:28 – 1:23:54
Creativity, ‘universal explainers,’ and the limits of explanation
Dwarkesh connects Gödel, creativity, and Deutsch’s ‘universal explainer’ thesis. Aaronson resists categorical claims: he doubts there’s a single ‘algorithm for creativity,’ acknowledges qualitative thresholds between humans and other animals (language, writing, universal machines), but warns that some questions may resist what we currently count as explanation, analogizing to fixed independence phenomena like busy beaver.
- •‘Algorithm for creativity’ as near-oxymoron; outputs would feel non-creative
- •Gradient of general reasoning: ant → chimp → human → Einstein
- •Human thresholds: recursive language, writing, number systems, computers
- •Skepticism toward Deutsch’s black-and-white optimism and definitions
- •Busy beaver as a model for fixed limits: some truths may outstrip current explanatory frameworks
- 1:23:54 – 1:27:05
Advice to young technical learners: exploit resources and become expert fast (narrowly)
Aaronson closes with practical guidance: learn aggressively using abundant modern resources, read current papers, and talk to authors. His core tactic is to pick an open, narrowly scoped problem, become the world expert on it, and let that success create a path to broader expertise and collaborations.
- •Never-before level of accessible learning resources (courses, books, arXiv)
- •Read new papers routinely; mine them for open questions
- •Engage researchers directly; ask for feedback and calibration
- •Become the world expert on one tiny problem first
- •Use that foothold to expand scope and form collaborations
