Stephen Wolfram: Cellular Automata, Computation, and Physics | Lex Fridman Podcast #89

The Fermi Paradox asks why, if extraterrestrial intelligence is likely, we haven't seen any evidence of it. Wolfram speculates that humanity might be encountering two forms of alien intelligence simultaneously: artificial intelligence (our own creation) and extraterrestrial intelligence. He suggests people will become comfortable with the idea that both exist and are 'intelligent like we are,' even if they don't share our civilizational history.

Wolfram's Principle of Computational Equivalence states that almost all processes that are not obviously simple correspond to computations of equivalent sophistication. This means that even systems with very simple rules can quickly reach the same level of computational sophistication as more complex systems, including universal computation.

Wolfram suggests that the universe itself might be computational in this sense, emerging from a simple program or rule. If this is true, then understanding physics would become a problem of solving complex, irreducibly difficult mathematical problems within the framework of that underlying program, rather than discovering new fundamental laws. He is optimistic that a fundamental theory of physics could be discovered through computational exploration.

Computational irreducibility is a concept that arises from the Principle of Computational Equivalence. It states that for many complex systems, the only way to determine their future behavior is to run every step of the computation; there's no shortcut or 'smarter' way to jump ahead and predict the outcome. This implies that even human brains, being computational, cannot always predict complex systems better than the system running itself.

Wolfram is interested in the most 'structureless structures' that could underlie space and time. He proposes that physical space and time might be emergent phenomena from a network of hypergraphs, rather than fundamentally continuous. In this model, rules define how these hypergraphs transform, and what we perceive as space and time emerges from their collective behavior.

Wolfram suggests that a property called 'causal invariance' in hypergraph rewriting rules could imply special relativity. Causal invariance means that the order of microscopic rewrites does not affect the causal network of events. This independence of microscopic order is mathematically analogous to the independence of physical laws from different reference frames in special relativity, allowing for a derivation of relativity from an underlying theory.

The book generalizes the idea of using mathematical equations to describe the world to using simple computational rules, like those found in cellular automata. The main insight is that even extremely simple rules can generate immensely complex and seemingly random behavior, suggesting that nature's complexity might arise from very simple underlying principles rather than complex ones.

Rule 30 is a specific one-dimensional cellular automaton that produces highly complex, seemingly random patterns from a very simple starting condition (a single black cell). Its significance lies in demonstrating that profound complexity can emerge from elemental rules, challenging the intuition that complex outcomes require complex initial conditions or rules. Wolfram used it extensively for generating random numbers.

Wolfram has offered $10,000 prizes for solving three problems: 1) proving whether the center column of Rule 30's pattern ever repeats, 2) determining if there's an equal number of black and white cells down the middle column over finite or infinite steps, and 3) finding a way to predict a cell's color at a given position/time faster than simulating every step.

Wolfram Language is a high-level symbolic computational language designed to express computational thinking and world knowledge in a way understandable by both humans and machines. Its goal is to allow direct computation with concepts from various fields, from cities and chemicals to algorithms, by having 6,000 primitive functions that cover a broad range of real-world entities and operations.

Wolfram Alpha acts as a question-answering system that takes natural language queries, converts them into Wolfram Language, and then uses its vast underlying knowledge base and computational capabilities to generate reports and answers. The knowledge base is a continuously growing collection of curated, computable facts about the world, allowing it to perform calculations and deliver precise answers.

Wolfram believes that as more of the world is automated, encoding ethics computationally becomes necessary. Regarding internet content selection (e.g., social media feeds), he suggests that there shouldn't be one universal 'AI ethics module.' Instead, a 'market-based' system where individuals can choose their preferred ethical framework (e.g., conservative, libertarian) for content filtering might be more appropriate, particularly for personal content selection.

Wolfram defines intelligence, in human terms, as 'computation that is doing things we care about.' He argues that in a broader, general sense, intelligence is simply computation, and that many simple computational systems can achieve comparable sophistication to what humans consider intelligent. The 'special' aspect of human intelligence is its connection to our goals and civilizational history.

Yes, Wolfram believes consciousness can emerge from computation. He finds the concept of 'subjective experience' deeply slippery and suggests it's another term that lacks a clear, ground-truth definition apart from computation itself. He cites discussions at AI ethics conferences about AI rights, where the emergence of consciousness in AI is often posited as a prerequisite and could lead to AIs demanding rights.