Time Doesn't Exist, and the Universe Didn't Come From Nothing
Cosmologist Brian Keating makes a counterintuitive case: time may not exist, and the universe did not come from nothing. He also warns that AI cannot reproduce human intuition, and cannot be fully controlled.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
Time may make no sense because it does not exist at all
Keating argues that physics can explain space, matter and energy, yet does not understand why time behaves the way it does — and the reason may be that time does not exist at all. He cites Frank Wilczek's definition, ‘time is what clocks measure’, and criticizes it as a tautology. The question he is actually working on is ‘what happened on the Sunday before the Big Bang’ — that is, how time itself arises out of nothing.
— Brian KeatingThe universe didn't come from nothing: equations don't instantiate things
Keating flatly rejects Lawrence Krauss's argument that the universe came from nothing as nonsense. Equations describe things; they do not instantiate things. He pushes on what came before the Big Bang — possibly a prior universe, possibly an eternal energy field, but certainly not ‘nothing’. That sentence is the key to his whole argument.
— Brian KeatingWith the Antarctic array cut, only Chile still runs this experiment
Because the U.S. government canceled the BICEP Array in Antarctica, the Simons Observatory at 5200 meters in Chile has become the only project of its kind still taking data. It produces roughly 1TB of data per day, is analyzed by 450 participating scientists, and is the first such experiment powered by solar panels. Keating's team is trying to use it to make a decisive measurement of the cosmological time field.
— Brian KeatingDetect primordial gravitational waves and every rival model is out
If primordial gravitational waves are detected, it would strongly support inflation while ruling out every other model — an eternal universe, a cyclic one, one that ultimately evaporates, or one sitting at the center of a black hole. Keating stresses this would not be final proof, but like Newton's laws being good enough to guide a Moon landing, that kind of evidence is still extremely powerful.
— Brian KeatingA congresswoman says non-human biologics exist but never shows evidence
Keating criticizes the extremely poor information environment around UAP. Even with authoritative figures such as Congresswoman Luna claiming that non-human biologics exist, no physical evidence has ever been produced. He argues that if real evidence existed it should have been made public long ago — otherwise it is a waste of the public's attention.
— Brian KeatingAI has been trained into an apology machine that internalizes self-negation
Responding to Peter's worry about distrust in nations and in science, Keating argues that Britain contributed enormously to the world, and yet British people are leaving. He criticizes judging Churchill and Newton by today's standards, and notes that AI has already been trained into an ‘apology machine’ that refuses to answer certain physics questions. He argues we have to stop apologizing for human greatness, or AI will internalize that self-negation.
— Brian KeatingLLMs are just matrix multiplication; they could never invent chess
Keating explains that LLMs rest on matrix multiplication — linear algebra, not advanced mathematics — and that GPUs were originally optimized for gaming, not designed for superintelligence. His example: an LLM could not invent chess or the Rubik's Cube; it can only work from data humans already produced. He invokes Jevons paradox: the wider AI is used, the more costs fall but total usage rises. LLMs cannot train on themselves, or they degrade like mad cow disease; they need new human data.
— Brian KeatingThe real risk isn't AI escaping control, it's that we already can't live without it
Keating raises two main worries: we do not understand how AI works, and we can neither predict nor control it. He brings up Turing's halting problem as an analogy for the unpredictability of modern AI. What worries him more is that if the AI bubble bursts — with OpenAI losing 10 billion dollars a month — humanity will be paralyzed by how dependent it has become.
— Brian KeatingIn their own words · checked verbatim
We understand space, we understand matter, we understand energy. We don't understand why time works the way it does. Actually, the reason is because there is no time. Time doesn't exist.
Brian Keating0:00
Something can't truly come from nothing.
Brian Keating18:55
It's not time that's limited. It's attention.
Brian Keating1:01:54
I don't believe in gravity. I have evidence for gravity.
Brian Keating1:10:05
We have to stop apologizing for humanity's greatness.
Brian Keating1:16:11
He said, that was the happiest moment of my life.
Brian Keating1:23:21
The only way of knowing the limits of the possible is to go beyond them into the impossible.
Brian Keating1:38:55
Figures
| Current average temperature of the universe | 2.726 Kelvin above absolute zero | 27:08 |
| Simons Observatory altitude | 5200 meters | 41:27 |
| Simons daily data volume | about 1TB | 41:27 |
| Scientists participating in Simons | 450 | 41:27 |
| OpenAI monthly losses | 10 billion dollars | 1:29:41 |
| OpenAI employee stock valuation | 800 billion dollars | 1:29:41 |
| Age of the meteorite | 4.3 billion years | 1:33:44 |
| Age of the Earth | 4.1 billion years | 1:33:44 |
Glossary
- BICEP Array
- A cosmic microwave background polarization experiment in Antarctica, built to look for the imprint of primordial gravitational waves left by inflation.
- Simons Observatory
- A cosmology experiment on Chile's Atacama plateau, measuring the polarization and temperature of the cosmic microwave background.
- Jevons Paradox
- Technological progress raises efficiency, and can nonetheless increase the total consumption of a resource.
- UAP
- The official term used in place of UFO, covering unidentifiable aerial and trans-medium phenomena.
- Cosmological Time Field
- A hypothesized field that encodes the rate at which time flows and could be observed, providing a mechanism for the arrow of time.
How to listen
Tech founders, AI engineers, science skeptics, and anyone who wants half an hour on cosmology and where AI hits its limits.
If you only care about AI, skip 27:08–35:14 on the early universe and how the days of the week got their names.