Knowledge Creators Outlive Stars, So Epistemology Is Physics
Twist quantum mechanics, computation, evolution and epistemology into a single strand, and the conclusion is: as long as humans keep solving problems, the long-term structure of the universe will be determined by knowledge, not gravity.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
Unpredictable does not mean non-deterministic
Brett Hall first dismantles a common misunderstanding: physics determining everything does not mean physics can predict everything. He gives three kinds of unpredictability: a system so complex that any computer that could be built cannot outcompute the universe itself; a single photon on a half-silvered mirror, which can only yield a determinate 50-50 probability; and, most crucially, the creation of knowledge. The laws of physics permit knowledge to be created, but the same laws forbid us from predicting which path that knowledge will take, because that would amount to possessing the knowledge before it is created. So ‘physical determinism’ and ‘a predictable future’ are two different things, and conflating them leads people to think epistemology will sooner or later be swallowed by physics.
— Brett HallProof requires matter, so proof can go wrong
In the original text Deutsch admits he changed his mind: he once thought human knowledge would unify under ‘ever-deeper fundamental physics’, and later rejected this. Brett unfolds the reasoning: a mathematical proof requires a person or a machine to carry it out, the thing doing the proof is made of matter, matter obeys quantum physics, and quantum physics means the possibility of error can never be ruled out. So a proof is never a proof of ‘necessarily true’, and mathematical knowledge is conjectural too. He calls the belief that mathematicians can intuit certain, error-free truths the mathematician's misconception, and quotes Deutsch: necessary truth is merely the subject matter of mathematics, not the reward of doing it.
— Brett Hall‘This theory could be wrong’ is a vacuous remark
Brett treats this sentence as his biggest bugbear when explaining this kind of content: there is always someone saying ‘but this theory could be wrong’, as if no one had thought of it. He concedes it is true, trivially true, part of fallibilism — but ‘could be wrong’ by itself solves nothing. The right approach is to take the current best explanation seriously and see where it takes you; if you assume from the start that it is wrong, then anything goes, and what is left is supernaturalism, magic and wizards. He gives ‘the Earth is approximately spherical’ and ‘energy is conserved’ as examples — these could also be wrong, but you cannot refuse to use them on that account.
— Brett HallOnce proved, the Turing conjecture is upgraded to a principle
Deutsch's 1985 paper proved the Turing conjecture, so many people now call it the Turing-Deutsch principle. Brett explains the difference between conjecture and principle: a conjecture is a guess that looks reasonable and that no one has a deep criticism of; a principle is a law about laws, one level deeper. If quantum theory is the total explanation of matter and energy, and from this conjecture one can prove that the Church-Turing conjecture holds, then it deserves the title of ‘principle’ — any law discovered in the future within quantum theory must conform to it, or quantum theory itself would have to change. Deutsch also stresses that this principle has an emergent character, bearing directly on the properties of complex machines and only indirectly on subatomic processes, with a status comparable to energy conservation and the laws of thermodynamics.
— Brett HallOn long timescales, the universe's structure is decided by knowledge
Following Deutsch's hint, Brett pushes to a picture: on the largest timescales, the structure of the multiverse will increasingly be determined by people and the knowledge they create. What people build and control in different universes will outlast any single star; knowledge across different universes will converge, because everyone is exploiting the same set of physical laws. He describes explanatory knowledge and its technological consequences as ‘like a crystal that persists’. So epistemology is the most important discipline on the long-term scale of cosmic evolution — this sounds very strange the first time you hear it, but its only premise is: humans keep solving problems and never go extinct.
— Brett HallTreating one strand as the whole world leads you astray
Deutsch says that quantum physics swallowing the other three strands is merely ‘the parochial view of a physicist’, possibly contaminated by reductionism. Brett gives three examples of treating a single strand as a worldview: computer-science types who see everything through the lens of deterministic computation, who think that given enough compute, hardware will ratchet itself up to AGI, then ASI, and take over the world, and who even think we are already living in a computer simulation; Brett Weinstein and others who want to explain all human activity, including politics, religion and social movements, by biological evolution; and Cartesian epistemological foundationalism. His judgment: each strand on its own is rich enough to support an entire worldview, but doing so is both limiting and misleading.
— Brett HallDescartes and the Buddhists arrived at the same place
Brett points out that cogito ergo sum does not appear where the full argument is given in the Meditations; what is said there is ‘that I think and that I exist is a necessary truth’, resting on ‘clear and distinct ideas’. He finds this reads as very Buddhist: Descartes sits in his armchair by the fire, uses the method of doubt all the way down to doubting the body, and finally finds there must be something doing the doubting. Buddhists, modern meditators and Descartes, each by their own method — turning inward to observe subjective experience — arrived at the same conclusion. Brett says this may not be a coincidence, because they come from the same place.
— Brett HallIf the simulation hypothesis were true, it would expose itself
Brett relays Peter Boghossian's challenge: even if the simulation hypothesis solves no problem, it could still be true. Brett disagrees, on the grounds that ‘problems are soluble’: if the universe is a simulation, there must be consequences that show up in our reality and are eventually discovered by us. He gives a concrete mechanism — the computer a simulation runs on has finite memory and speed, so physical constants may stop gaining digits once measured to some precision. He takes the gravitational constant G as an example: it is now 6.6743 times 10 to the minus 11, and as instruments get more precise the digits should keep increasing; if one day a measurement reads 6.6743108 followed by all zeros and never changes again, that would need an explanation, and the simulation hypothesis would explain it. But right now there is no reason at all to think it is true.
— Brett HallPredicting stars means factoring in human choices first
Brett explains Deutsch's precise use of ‘fundamental’: a thing is fundamental if it is caught up in the explanation of many things — gravity, for instance, is used to explain tides, falling leaves, the Moon's motion, stars shining and the evolution of the universe as a whole. So why is epistemology fundamental? Because on cosmological timescales — thousands, millions, billions of years — predicting the far future requires first considering what the people alive then will do. A star would, on astrophysical models, expand, cool, contract, reheat and cool again; but if there are people there, with knowledge, resources and wealth, they may choose to slow down or speed up the star's evolution and reactions. Human choices, economics and preferences enter the prediction, so epistemology is not only about belief and subjective psychology, but also about stars, galaxies and the universe.
— Brett HallIn their own words · checked verbatim
the most important kind of unpredictability is that of knowledge creation. Physics, deterministic physics, physics that consists of... regularities of nature or laws allows for knowledge creation. But those same laws seem to prohibit our predicting of the route knowledge creation will take.
Brett Hall4:24
necessary truth is merely the subject matter of mathematics, not the reward we get for doing mathematics
Brett Hall8:29
In the meantime, it doesn't get you anywhere. In the meantime, it doesn't allow you to solve problems. In the meantime, what you should be doing is taking the best explanation seriously and see where it leads.
Brett Hall11:32
the structure of the multiverse will be increasingly dictated by people and the knowledge they create. Those structures will persist longer than individual stars
Brett Hall14:32
It's exciting to create knowledge and it's a dead end to claim that you're absolutely certain and without doubt. And it's completely unnecessary.
Brett Hall26:44
investigating fantasies, and they are fantasies, seems like a waste of cognition
Brett Hall34:59
We can't predict what a star will do in billions of years unless we take into account the choices of people that might be there at that time.
Brett Hall41:09
Figures
| Current measured value of the gravitational constant G | 6.6743 × 10⁻¹¹ m³/(kg·s²) | 31:52 |
| Year Deutsch proved the Turing conjecture | 1985 | 12:32 |
Glossary
- Turing-Deutsch principle
- A physical principle upgraded from the Turing conjecture after Deutsch proved it, constraining every information-processing system.
- mathematician's misconception
- The belief that mathematicians can intuit certain, error-free, infallible truths.
- Omega Point Theory
- A hypothesis about the endpoint of the universe proposed by the physicist Frank Tipler; Deutsch considers its core valuable but exaggerated.
- crypto inductivist
- A critic of Popper's epistemology for describing only the conditions under which scientific knowledge grows, not why it grows.
How to listen
Suited to people who have read or are reading The Fabric of Reality and want to understand how Deutsch elevates epistemology to the status of a physical principle; also suited to engineers who have their own views on the simulation hypothesis and AGI determinism.
18:37–21:39, the digression on Brett Weinstein and group selection, is loosely connected to the main thread.