Wolfram, Stephen
▶ Watch on YouTubeCore ideas
Wolfram traces his intellectual arc from his 1980s discovery that even the simplest possible rules — cellular automata like Rule 30 — can generate irreducibly complex, seemingly random behavior, a finding he says overturned his own initial assumption (from physics) that complexity requires complicated underlying causes. This led him to the concept of computational irreducibility: for many systems there is no shortcut to knowing the outcome of a process other than actually running it step by step, which he argues is a fundamental limit on prediction (illustrated by his standing "Rule 30 prize" for basic questions about the rule that remain unresolved) and which he suggests may make apparent randomness in nature not fundamental but simply a reflection of an observer's own computational boundedness relative to an underlying deterministic process.
The bulk of the conversation lays out the Wolfram Physics Project's model of fundamental physics: space is not a continuum but a network of discrete "atoms of space" connected by relations (hyperedges of a hypergraph), with roughly 10^400 such atoms comprising the observable universe and an elementary length scale around 10^-100 meters; time is not a fourth dimension alongside space but is identified with the process of applying rewriting rules to this hypergraph, step by step. Wolfram explains how requiring "causal invariance" (that the final causal structure of events doesn't depend on the order in which applicable rewrite rules are applied) is what forces the model to reproduce relativity, and how the same hypergraph-rewriting framework naturally generates General Relativity's field equations from purely combinatorial, rule-based dynamics — a result he considers among the project's most striking, since it wasn't built in by design but falls out of the formalism.
Quantum mechanics, in Wolfram's account, emerges from taking seriously that there are multiple ways to apply the underlying rules at each step, producing branching "multiway" histories rather than a single thread of time; observers are literally "branching brains perceiving a branching universe." He introduces "branchial space" — an abstraction of the space of these quantum branches, analogous to physical space but built from branch-adjacency rather than spatial adjacency — and uses it to explain destructive interference (two paths ending up maximally separated in branchial space) and to propose a mechanistic account of time dilation: motion, in this model, consumes some of the same computational "effort" that would otherwise go toward advancing an object's internal clock, so a moving object's clock runs slow because part of its computation is diverted into "recreating itself" at a new spatial position. He extends the analogy further, suggesting a maximum "entanglement speed" in branchial space parallel to the speed of light in physical space, and floats possible observational signatures of the theory — fluctuations in effective dimension in the very early universe potentially visible in the cosmic microwave background, or entanglement-speed effects detectable via black hole mergers or many-body quantum systems.
The conversation's most speculative arc concerns why the universe exists at all. Wolfram's answer centers on the "Ruliad" — the entangled object formed by running every possible computational rule in every possible way, forever. He argues the Ruliad is not a contingent choice (unlike picking a specific rule for "our" universe) but a necessary object, in the same sense that 2+2=4 is a necessary consequence of definitions rather than an arbitrary fact — and that any observer embedded in the Ruliad, merely by virtue of being computationally bounded and experiencing a single thread of time, will inevitably parse out something that looks like our known physical laws. He explicitly flags this as one of his least fully worked-out ideas (acknowledging a "hyper-Ruliad," incorporating non-computational oracle rules, that he says he hasn't thought through) and connects it to a parallel research program in "metamathematics" — treating mathematical statements and proofs as a space to be explored the way physical space is, complete with its own analogues of black holes (decidable theories, where "time" — proof search — necessarily terminates), time dilation (the cost of translating between mathematical fields to prove the same theorem), and possibly higher-categorical "topology" of proof space. He describes wanting to extend this "multicomputational" paradigm — systems with many simultaneous asynchronous threads of computation, parsed differently by different kinds of observers — well beyond physics, into chemistry (asking whether chemical reaction networks encode information in their dynamics, not just in molecular structure, with implications for molecular computing and immunology), biology, economics (modeling transactions as fundamental events and asking whether economic law has a General-Relativity-like global theory), and distributed systems and blockchain (his company's collaboration with Cardano, computational contracts, and speculative "physics-inspired" distributed ledgers).
He closes with reflections on his own career pattern of alternating between basic science and technology roughly five times, framing his life's projects — Mathematica/Wolfram Language, A New Kind of Science, and the physics project — as "artifacts from the future" whose value is sometimes only recognized decades later, and expresses regret that the 1980s complexity science movement he helped found got diluted by premature rushing into applications rather than developing what he now calls "metamodeling" (finding the computational primitives underlying diverse models) and "ruliology" (the basic, application-agnostic study of what simple computational rules actually do) as disciplines in their own right, comparing the latter to a timeless, foundational activity akin to pure mathematics.
Wolfram frames consciousness and intelligence as related but distinct: he defines "intelligence" broadly as computational sophistication (which by his Principle of Computational Equivalence is generically as sophisticated as anything else once a low complexity threshold is crossed), while consciousness is a narrower notion requiring specific additional constraints — computational boundedness and a single, sequential thread of subjective time/experience. This leads him to speculate that alien or artificial intelligences occupying different "reference frames" in the abstract space of possible rules (Rulial space) could perceive the same underlying Ruliad in ways potentially unrecognizable to humans, much as human observers are bound to a particular location and perspective in ordinary physical space.
Excerpts
On computational irreducibility as a hard limit on prediction:
“The only way to find out the answer to that computation is essentially to run it and just watch what happens... there's no way to shortcut it, there's no way to jump ahead and say I can figure out what's going to happen without going through all the computational steps.”
On why the Ruliad is, in his view, a necessary rather than contingent object:
“There's no choice being made — there's no 'oh, we picked this particular universe and not that one'... this Ruliad object is in a sense a necessary object, it is just a thing that is the consequence of working out the formal definition of things... it is not a thing where you're saying this is picked as the particular thing, this is just something which necessarily is that thing because of the definition of what it means to have computation.”
On the mechanistic explanation of time dilation as a computational trade-off:
“As you try to move, in effect some of the activity that would have gone into, so to speak, computing what happens to the clock is diverted into recreating you in a different place... and that's kind of the essence of time dilation in this model.”