Cronin, Lee

chemist, University of Glasgow, creator of assembly theory

Lee Cronin: Controversial Nature Paper on Evolution of Life and Universe

Lex Fridman Podcast · Dec 9, 2023

• Lee Cronin
• assembly theory
• origin of life
• complexity and selection
• Nature paper controversy
• astrobiology and biosignatures
• time and free will
• Kolmogorov complexity
• artificial general intelligence skepticism
• Chem Machina

Core ideas

Cronin explains assembly theory, the subject of his controversial Nature paper "Assembly Theory Explains and Quantifies Selection and Evolution," as a framework for quantifying the complexity of any finite, decomposable object by finding the minimum number of steps ("assembly index") required to build it from elementary parts, reusing previously-built substructures rather than starting from scratch each time (illustrated with Lego bricks and molecular bonds). Critically, high assembly index alone doesn't distinguish a truly complex, selection-produced object from mere randomness — what does is copy number: finding many identical high-complexity copies of an object is strong evidence some non-random "factory" (a cell, an environment, a causal process) had to have built it repeatedly, since a random process essentially never produces multiple identical highly complex objects by chance. He walks through the theory's four nested "universes" (assembly universe: anything combinatorially possible with no physical constraints; assembly possible: constrained by physical laws but with instantaneous access to any construction path; assembly contingent: objects can only be built by traversing an actual causal chain, with no shortcuts; assembly observed: what we actually detect), arguing the contingent-vs-possible distinction is the theory's central claim — that the universe does not grant instantaneous access to highly assembled structures, and something must physically encode the selection process that got there.

He describes practical validation: using mass spectrometry, infrared spectroscopy, and NMR on molecules to independently measure assembly index (all three techniques converge on similar values), and testing the approach blind on NASA-supplied samples (including ancient fossils and a Murchison meteorite) where objects above an assembly index of roughly 15 reliably corresponded to biological origin and objects below it to abiotic origin — which he proposes as a general "life detector" methodology for astrobiology: rather than searching for Earth-specific biomarkers, look for molecules with both high structural complexity and abundant identical copies, a signature he argues is universal and independent of any particular biochemistry. He also describes newer, unpublished work using assembly theory (via mass spec fingerprints of coexisting molecules rather than gene sequencing) to reconstruct the tree of life, and speculates about extending the framework to language, cellular automata rules, drug discovery (focusing on ligand-evolution rather than exhaustive protein-target modeling), and eventually culture and technology.

On the fierce backlash to the paper (2.3 million Twitter engagements, hundreds of thousands of downloads), Cronin recounts that it managed to simultaneously anger evolutionary biologists (who felt the origin-of-life question was already "solved" or that the paper overclaimed ownership of "selection" and "evolution"), physicists (skeptical of claims about initial conditions), prebiotic chemists (wanting more chemistry), computational-complexity researchers (who saw overlap with Kolmogorov complexity), and even creationists (who found it "not creationist enough") — which he treats as evidence the paper struck a genuine nerve rather than as a mark against it. He distinguishes assembly theory from Kolmogorov/algorithmic complexity by arguing the latter ignores causation entirely (a compressed description says nothing about the causal history that produced an object), whereas assembly theory is fundamentally about inferring a real, physically-instantiated construction pathway. He also recounts the earlier, harsher rejection of a related mass-spectrometry paper (six rounds of review, a reviewer accusing the team of fraud) as formative, and frames sustained, specific, "actionable" criticism — as opposed to vague dismissal — as something he actively seeks out and values, tracing this attitude back to being placed in a remedial learning-difficulties class as a child due to poor reading and math scores, and gradually proving capable through persistence rather than early promise.

A large section of the conversation is devoted to Cronin's claim that time must be fundamental (not an illusion or coordinate) and that the universe is "deterministic looking backward, but undetermined going forward" — because, he argues, no finite object (or the universe itself) can store the infinite precision needed to make classical or quantum mechanics genuinely deterministic into the future, and because "the future is bigger than the past," meaning the universe cannot contain its own future the way block-universe or multiverse pictures assume. He connects this directly to free will (arguing that if time is not fundamental, free will is impossible) and to a striking claim that Earth — via its unbroken causal/combinatorial lineage from LUCA through multicellularity, culture, and technology — is "the biggest thing in the universe we know of," while speculating that even if life is common throughout the cosmos, each planet's combinatorial history may be so different that humanity may never be able to communicate with alien intelligence even if it exists nearby, which is part of his motivation for trying to synthesize alien-like life in the lab as a comparison case.

On AI, Cronin is a self-described "AGI is far away" skeptic and pushes back hard on doomer arguments (deliberately relabeling AI "autonomous informatics"), arguing that current large language models are fundamentally interpolating from a training-set past rather than generating true novelty, that intention and reasoning in these systems ultimately trace back causally to human prompters and trainers rather than existing in the model itself, and that comparisons to human-level or superhuman intelligence are a category error since the only working examples of general intelligence come from biological evolution. He argues probability estimates of AI-driven extinction ("p(doom)") are unfounded because, unlike an asteroid impact, there is no known causal mechanism to estimate a probability from, comparing AI-doom reasoning to worrying about hypothetical "anti-gravity" — while conceding that real, present-day harms (fake data, fake users, manipulation) deserve serious concern and regulation. He describes his own "Chem Machina" project, attempting to build a chemistry-based (rather than silicon-based) substrate for intelligence, arguing that if genuine novelty-generation is ever demonstrated by an artificial system — something not reducible to its training data — that would change his view.

The conversation closes with a wide-ranging, half-serious discussion of nuclear weapons policy (a thought experiment about widely distributing weapons to deter their use, later reframed as running deterrence simulations in a "metaverse" instead) and Cronin's views on God and meaning: he identifies as an atheist but not a dismissive one, proposes that "selection" itself is the closest thing to a creative force in the universe, and frames free will as operating specifically at the boundary between a determined past and an undetermined future — with imagination itself treated as having genuine causal power on what comes next, which he says is the most exciting, unresolved mystery he's chasing.

Excerpts