Every Atom in Your Body Was Forged in a Dead Star

Lex Fridman opens with Rees’s 2020 *Scientific American* line: the universe is far bigger and stranger than anyone suspected. What we have learned, Rees says, is an origin about 13.8 billion years ago in a hot dense Big Bang whose very beginning is still shrouded, plus the zoo of extremes—black holes, neutron stars, explosions—and, in the last couple of decades, the realization that most stars have retinues of planets. The night sky is no longer points of light. It is planetary systems, which raises the 21st century’s exciting question: is there life out there? Life could be unique to Earth, or it could be routine given young-Earth conditions, in which case our galaxy hosts billions of biospheres. Settling which extreme is true is a job for the coming decades.

Rees will not give a firm estimate of whether we are alone, because we do not know, and because biology is harder than physics. Most of the known universe can be understood with physics. Even the smallest living thing, an insect, is far more complicated, layer on layer, than the most complicated star or galaxy. Fridman asks the physicist’s dream question: if we unlock how the universe works, do we then understand how life emerges from that fabric? Rees calls the phrase theory of everything very misleading. It names a unification of electromagnetism, the weak interaction, and gravity. That is an important step for particle physicists. Its absence does not hold up anyone else.

Biologists are stuck on complexity, not on subnuclear ignorance. Extreme reductionists like to say we are all solutions to Schrödinger’s equation. Rees does not believe in a special life force, but nobody thinks you understand an organism by solving that equation. A breaking wave is a solution for 10^30 particles. Even if you could solve it, you would get no insight into viscosity or turbulence. Every science has its own irreducible concepts: valence in chemistry, imprinting in ecology, others in psychology. Sciences stack like a tall building, physics in the basement and economists in the penthouse, but an unstable base does not make the upper floors wobble. A chemist’s problems are not made worse by uncertainty about quarks.

If you care how birds fly, you need imprinting, not an atomic-scale solution of an albatross wandering thousands of miles over the Southern Ocean and coughing food up for its young. You could not write that equation even in principle. Understanding, Rees says, lives at the level where the concepts actually explain.

There is no particular reason, Rees says, that brains evolved on the African savannah should match the deepest structure of reality. A monkey cannot understand quantum theory or even Newton. Some aspects of the world we may not even be aware we fail to grasp. The interesting intermediate is AI. String theory’s ten-dimensional geometry may be too hard for a human to work through and still be tractable for a machine that can learn world-championship chess in a few hours by watching games. If that is how we come to trust the theory, it would be frustrating: you would not get the aha that is science’s greatest satisfaction. If the machine then spits out the electron’s mass and the fact of three neutrinos, you would know there was truth in it. You would trust a theory because it predicts what you can check, without ever understanding its workings the way you more or less understand everyday phenomena.

Fridman spent a day at DeepMind. Rees knows Demis Hassabis well and calls him one of his heroes. The first time they met, Hassabis said he wanted to understand the universe too, but thought the best route was to develop AI and then use it. Rees thinks he is right. AlphaFold already understands protein folding better than the best human chemists. DeepMind is also suppressing instabilities in tokamak plasmas better than any other method. Computers win by chewing enormous samples fast: they can look at 100,000 lung scans in a few days, more than a radiologist sees in a lifetime. Asked whether an AI will win a Nobel this century, Rees plugs his next book, *If Science Is to Save Us*: a chapter arguing that Nobel Prizes do more harm than good, because they distort how the public thinks science actually gets done.

Astronomy’s huge advances, he says, did not come from being smarter than Aristotle. They came from better data—telescopes on the ground and in space—and from simulations. A particle physicist can smash beams at CERN. Rees cannot smash two galaxies together, but in a computer he can tweak initial conditions and see which run fits what we observe. The European Gaia satellite has put online the speeds, colors, and properties of nearly two billion Milky Way stars. Machine learning loves that kind of organized raw data. Human intuition still has magic, Fridman hopes. “We hope so. For now,” Rees replies.

More than twenty years ago Rees guessed we would have solved dark matter by now. We have not. Galaxies would fly apart if they contained only what shines. To keep them the size they are, and to form them in the first place, you need about five times as much gravitational stuff as the gas and stars we see. It is not dark so much as transparent: a swarm of microscopic particles with no electric charge and a tiny cross-section, felt only in bulk. Simulations of galaxy formation work better if you put in non-interacting particles rather than a gas that would shock. The Large Hadron Collider was supposed to make the obvious candidates and did not. Some people say dark matter cannot be there. Rees disagrees. Axions remain good candidates, and there are ten powers of ten between the heaviest particle the LHC can make and the heaviest that could exist without collapsing into a black hole. We have explored a sliver of parameter space.

The bigger gap is the ultra-early universe. From a nanosecond onward, the physics is in the laboratory range: particles bouncing like those in the LHC, and after a second, like the cores of the hottest stars, cooking hydrogen and helium in abundances that fit the data. Winding the cosmos back to a nanosecond old can be done with about as much confidence as a geologist talking about Earth’s early history. That is huge progress in fifty years. Progress sharpens new mysteries: why the expansion rate, why this mix of atoms, dark matter, and radiation, why smooth enough for galaxies but not perfectly smooth. Those answers, he believes, lie far earlier, when conditions were beyond any experiment.

How big is the gap? The observable universe, a bit more than ten billion light-years, squeezed at one nanosecond into the volume of the Solar System. The era when the key properties were imprinted squeezed that entire universe to the size of a tennis ball, or a baseball if you prefer, emerging from something microscopic. Rees hopes for a theory that applies in that extreme and also leaves traces we can test now—why neutrinos exist, or the unexplained numbers in the Standard Model. AI may have to help with the calculation. Until then it is speculation, not a fork into philosophy so much as science still groping.

What happened before the Big Bang is beyond present science, Rees says, but science’s frontiers move, and he sees no clean split with philosophy. Understanding that regime may demand concepts more counterintuitive than quantum theory. Brains have not changed much since ancestors roamed the savannah; it is already surprising we made any sense of the quantum world and the cosmos. Three dimensions of space and time ticking away are among the ideas you might have to jettison. Hawking’s line—that asking what came before is like asking what is north of the North Pole—is one option. Rees does not like it. Were there many Big Bangs? We may get a foothold from a new theory without being able to test every prediction.

That last claim is heresy to some, so he offers an analogy. We take Einstein seriously about the insides of black holes even though we cannot look, because the theory is vindicated by cosmology, gravitational waves, and the rest. If a theory explained our Big Bang and the present universe and then predicted many Big Bangs, we would take the inference seriously even if we never detect the others. Andrei Linde’s eternal inflation, on specific and unvindicated assumptions, produces new bangs forever. Other theories produce one. Rees wants open-mindedness, not dogma, until the physics is understood.

Even the most conservative astronomers, he says, would agree that physical reality is more extensive than the telescope. The observational horizon is about 15 billion light-years. That horizon is no more the edge of reality than the horizon around you on the ocean. There are quite strong arguments it goes on about 100 times further, perhaps far enough that every combinatorial is replicated and another set of people sit in a room like this. Look as far as you can one way and the opposite way and conditions differ by no more than one part in 100,000; if we are part of a finite structure, that tiny gradient says we occupy a small corner. Fridman pushes the metaphor: we also do not know the depth of the ocean, or the observing device in our heads. Rees agrees, then shrugs. Even the Solar System is vast by human standards. Humility follows if you started unhumble. Most of us, he thinks, are not much changed by the scale.

Fridman finds it almost obvious the universe should be full of alien life, given the vastness and how ordinary Earth looks. Rees does not agree, as far as life is concerned, because we do not understand how it began. The jump from complex chemistry to the first replicating, metabolizing entity is a mystery. Serious chemists are on it; we still cannot say whether it was a rare fluke or a process that happens on any young Earth. Even if simple life is common, Stephen Jay Gould’s point stands: rerun evolution and you may not get intelligence. These are open questions, which is why the moment is exciting. Origin of life used to sit in the too-difficult box. Clever experiments help. More important, we finally have places to look.

More than 5,000 exoplanets have been found, almost all indirectly: a transit dip, or a gravitational wobble of the parent star. Size and mass can be estimated. Actual light from the planet itself has barely been caught, except for a few huge, bright ones. James Webb may take steps. Better still will be ESO’s Extremely Large Telescope, a 39-meter mirror that can separate a nearby Earth-like planet’s light from a star millions of times brighter and read the spectrum for oxygen or chlorophyll. Imagine aliens staring at us: Carl Sagan’s pale blue dot, much fainter than the Sun. If they could observe that dot they could infer an atmosphere, the length of the day from Pacific versus Asian landmass, maybe seasons. That is the kind of inference we may draw in ten or twenty years.

The other path is closer. Something may swim under the ice of Europa and Enceladus. Probes will sample the spray for organics, even if they never drill through. If life originated independently twice in one planetary system, Rees’s logic is clean: it cannot have been a rare accident, and must have happened billions of times in the galaxy. Mars would still be ambiguous, because early life could have hitchhiked on meteorites either way. Skeptics ruin a good party, Fridman jokes. We need them at the party, Rees says. Intelligent aliens of science fiction are a different matter from vegetation under ice.

If life is everywhere, why is intelligence not? Rees chairs a committee funded by Yuri Milner at $10 million a year for better equipment and telescope time to search for intelligent life. He thinks it is worthwhile. He does not hold his breath. What intrigues him is what a detection would actually be. Nearly four billion years from first life to us; a flesh-and-blood technological civilization that can announce itself may last only a few hundred years more. Some people say it will kill itself off. Rees is more optimistic, and the optimism is strange: future evolution will not be slow Darwinian selection. It will be what he calls secular intelligent design—humans designing progeny better adapted to where they live.

Flesh and blood have limits, so those pioneers may go largely electronic, downloading brains. Electronic entities are near-immortal, need neither atmosphere nor gravity, and are not daunted by long interstellar trips. They drift into the blue yonder. If another planet ran Earth’s track with a head start, we would not see a synchronized flesh civilization. We would see those electronic progeny, still burping away far from home. The Fermi argument that they would have come and eaten us has an escape clause. Darwinian selection needs aggression and intelligence. Designed intelligence may keep the intelligence it was built for and drop the expansion. They may sit in deep space thinking deep thoughts, not grabbing territory. Competition, Fridman suggests, can live in the space of ideas. They would all be philosophers, perhaps, Rees says.

*The End of Astronauts*, co-written with Berkeley colleague Don Goldsmith, argues that the practical case for sending humans is getting weaker as robots get better. Fifty years ago robots could do little; now they can assemble large structures in space or on the Moon, dig on Mars, and, with future AI, do the geology. Europa and Enceladus are places you could never send people. A human to Mars needs feeding for 200 days and a ride home. Robots need neither. If humans go, it is as adventure. Rees does not want NASA or other public agencies to fund that, both because there is no practical purpose and because a public body has to be risk-averse. The shuttle flew 135 times and had two catastrophic failures, seven dead each time, including a schoolteacher on a flight sold as safe for civilians. National trauma, then more cost to make it safer.

Leave human spaceflight to billionaires and their sponsors, he says: taxpayers do not pay, and you can send people who will accept a 2% risk the way test pilots, hang-gliders, and polar explorers do. Space adventure, not space tourism. Cheer them on. Low Earth orbit will get cheap. Mars is expensive and dangerous; the only takers will be adventurers, maybe on one-way tickets like Magellan. He hopes that by century’s end a small community lives there, far less comfortably than at the South Pole, the ocean floor, or Everest, without a return ticket. Elon Musk’s talk of mass immigration to escape Earth’s problems is, Rees says, a dangerous delusion. It is far easier to deal with climate change here than to terraform Mars. There is no Planet B for ordinary risk-averse people.

The crazy pioneers are another matter. By century’s end, genetics and cyborg implants—even downloading—will have advanced, and those techniques will be tightly regulated on Earth for prudence and ethics. On Mars, away from the regulators, poorly adapted people have a motive we do not. They may become a post-human species, then electronic, then leave the planet altogether as precursors of whatever intelligence comes from Earth. Astronomers know what most educated people forget: Earth is 4.5 billion years old; the Sun has six billion years left before it engulfs the inner planets. Woody Allen’s line: eternity is very long, especially toward the end. We are not even a halfway stage in cosmic complexity. Whatever watches the Sun die will differ from us as we differ from slime mold.

Would those electronic tenants be conscious, or zombies that behave as if they were? Rees sees no empirical answer. Consciousness might emerge in any complicated network, or it might be special to flesh. After one talk, someone asked: wouldn’t it be sad if the universe’s main tenants had no self-awareness, nothing that could appreciate wonder and beauty as we do? That colors whether you welcome the scenario. Be humble, he adds: a strong ego tied to one brain may not be the highest form. Fridman suggests that from orbit, the atom of intelligence might be the collective—ideas jumping across generations—more than the person in the suit collecting prizes. Rees does not knock it down. Zombies may simply have a different way to experience.

Fridman pushes back with affection: reaching Mars inspires a generation, the way Apollo did, and Musk is not selling escape so much as a small-percentage backup against the self-destruction Rees himself writes about. Rees agrees the adventure inspires, and that billionaires who build rather than shuffle crypto are benefactors. His remaining point is cost under risk aversion. Democracies resist exposing people to high risk when we share responsibility. He credits SpaceX’s cheap reusable first stage—more than 100 tons to orbit, a nose cone that could swallow an unfolded Webb mirror, even space solar power no longer crazy—as a tremendous service. He is unlikely to see a boot on Mars in his own ancient lifetime. He still wants the century’s small community, and he still thinks the public will flinch at a no-return ticket, the way the shuttle’s 2% risk was unacceptable even after Apollo’s accepted gamble.

*2001* leads to HAL. Should a computer prioritize the mission over the crew? Rees uses the question to talk about everyday life. If a machine recommends an operation, denies parole, or refuses bank credit, you should be entitled to an explanation. It is not enough to hear that the machine’s track record is better on average. The trend toward decisions you cannot contest is dangerous. Use AI as a supplement: the radiologist has not seen as many cancerous lungs as the model, so let the model help, and let the human decide. On the labor market he wants heavier taxes on the conglomerates that deploy robots, funding decent jobs where being human is the point—carers for the old, classroom assistants, park gardeners—so people now in Amazon warehouses or call centers move to work where a person is an asset.

Long-term, Fridman says, biological *Homo sapiens* may not remain the default; the path from here to cyborgs will be painful, so slow the enhancement techniques on Earth and leave them to Mars. Rees agrees we are well adapted here and lack their motive. He also wants caution about exponential stories. Technologies spurt and level off. Apollo was twelve years after Sputnik; footprints on Mars in another ten or twenty seemed obvious, and that remains the high point of human exploration because it was superpower rivalry at huge public expense. Civil aviation jumped from the 1919 Atlantic crossing to the 1969 jumbo and then sat there. The iPhone 24 may not differ much from the 13. Concord came and went. Some advances that look unstoppable saturate, and not everything that makes money is good for society.

In *On the Future* Rees imagines a tweet-length message back to Newton. Newton would love planets around other stars, and that everything is atoms. If he looked more carefully through his prisms at flames, not just sunlight, he might have twigged that different substances emit different colors—clues that had to wait two or three hundred years. Until about 1850 astronomy was positions and motions; Newton understood ellipses but not why the Solar System sits in a plane, and nobody knew what stars were made of. They were a fifth essence, not earth, air, fire, and water. Only after precise spectra did people see that the Sun is the same stuff as Earth. Nuclear energy, and what keeps the Sun shining, arrived around 1930. Newton, given a quick course, might even have made sense of curved spacetime. He was a unique intellect who said he thought better by thinking on things continually. Darwin, by contrast, claimed only the common sense of an average lawyer: collect data, think it through. Science, Rees has learned over a lifetime among scientists, contains many mental styles.

Black holes are the century’s astonishing story. A year after general relativity, Schwarzschild found a solution that contained them. Einstein never took it seriously. Others did. Roger Penrose and colleagues in the 1960s showed that once a black hole forms and settles, it is defined by two quantities: mass and spin. The most violent objects in the cosmos become the most standardized. How they form is messy. The Sun burns hydrogen to helium and, in about ten billion years, shrinks to a white dwarf. Heavier stars burn faster, climb the periodic table from helium to carbon, then explode in an energy crisis and blow out the material. All the atoms in our bodies were synthesized in stars that lived and died more than five billion years ago. We each contain atoms from thousands of different stars across the Milky Way. Fred Hoyle suggested that in 1946. It has been borne out.

Massive stars leave two kinds of remnant. Neutron stars, found in 1967–68, pack more than the Sun’s mass into about ten miles—laboratories of physics we could never build. Still heavier cores cannot sit as a stable neutron star and become black holes of roughly 10 to 50 solar masses. The first evidence, fifty years ago, was a black hole orbiting an ordinary star, grabbing gas that swirled in and screamed in X-rays on a very short timescale, so the emitter had to be small and dense. Then came the monsters. A black hole at a galactic center, fed with gas, can outshine the other hundred billion stars: a quasar. They were found as star-like points that outshone galaxies. Many squirt jets across every waveband. The Milky Way’s central hole weighs about four million Suns—modest next to some measured in the billions—and is faint now because little is falling in. Most galaxies, Rees suspects, have one, and most went through a quasar phase, then went quiet.

Andromeda, a big disk with a black hole of its own, sits about three million light-years away in the same gravitational well and will meet us in about four billion years. Less a crash than a dance: they will pass through each other and merge. The two central holes will settle, whirl, and collide, ringing a huge burst of gravitational waves at about one cycle per hour, too low for LIGO, a job for a space interferometer. Descendants would be fine. Stars will not pack much tighter; the chance of a stellar collision with the Sun stays tiny. The Milky Way band across the sky will vanish because that is the disk seen edge-on, and the sky will hold twice as many stars in a more spherical spread. LIGO itself Rees ranks among the great engineering feats: precision like measuring a hair’s thickness at the distance of Alpha Centauri, or 10^−21. The credit, he says, belongs mostly to the engineers.

On cosmological timescales it will not be humans watching the Sun die. On human timescales this century is special. It may see the shift from purely flesh-and-blood entities to cyborgs. More urgently, it is the first century in which one species has the future of the planet in its hands. One class of threat is numbers and appetite: climate change and lost biodiversity on a scale that never happened because there were never this many humans. The other worries him more: misuse of technology. Small groups, even careless ones, can now create effects that cascade globally. His worst nightmare is a few people engineering a virus more transmissible or virulent than nature’s—gain-of-function work already done on flu a decade ago. COVID-19 showed that a disaster in one place cannot be confined. A 2012 Pentagon report on a state cyberattack against the eastern U.S. grid ended a chapter by saying it would merit a nuclear response. What needed a state actor then, he thinks, might now be done by a small group empowered by AI. Bio and cyber are the two biggest concerns. Imagine the pandemic without the internet.

The war in Ukraine shocked Fridman with how close nuclear use still sits: one mistake, one bad actor, one leader. Add cyber confusion and a launch could follow false information with no time for collective care. There have already been false alarms that were caught in time. Cambridge’s centre is looking at a threat that did not exist thirty years ago: more automated command and control of nuclear weapons and submarine fleets, open to cyberattack. Things happen faster; humans are less immediately in control; a local event can cascade globally as it never could. Interconnectedness raises living standards and cuts resilience. Long supply chains are efficient until one link stops car plants, as the pandemic showed. Filling 95% of intensive-care beds looks efficient; Germany’s habit of keeping 20% free is prudence. COVID, Rees thinks, taught a rebalancing of resilience against efficiency.

Even one person with university-lab gear making a dangerous pathogen is too many. Regulations are being written; enforcing them globally looks about as hopeful as enforcing tax or drug laws. Every nation, he says, faces a painful trade-off among freedom, security, and privacy. China will give up privacy for security, not necessarily for liberty. In Britain and America he thinks we will have to give up more privacy too. Fridman wants another defense: reduce the number of people who *want* to do this, by taking mental health, meaning, and flourishing seriously. Rees agrees, then points at Africa: almost everyone there has reason to feel embittered, cannot copy the East Asian tigers because robots took cheap manufacturing, and needs something like a mega Marshall Plan—not just European altruism, but self-interest. The 2,000 richest people hold enough wealth to double the income of the bottom billion. That, he says, is an indictment of the world’s ethics.

He and Steven Pinker have written together on biothreats. He accepts Pinker’s facts: life expectancy, infant mortality, literacy. He thinks Pinker misses the new class of threats, and he questions whether collective ethics have improved. In the Middle Ages the average life was more miserable, but the gap between how the world was and how it *could* have been was narrow, because knowledge was scarce. That gap is now far wider. Allowing it to stay wide makes us more ethically at fault than our predecessors. Social media, in his view, makes moderate consensus harder: the BBC and the papers used to muffle crazy extremes; now a click feeds you a more extreme next thing. Polarization will make every country harder to govern. Fridman says the platforms are revealing division, not only creating it. Rees’s reply: you are assuming the fat cats are more public-spirited than the politicians.

Getting politicians to care thirty years out is, Rees says, a matter of public opinion. Scientific advisers make the long-term points and lose to the urgent inbox. Enlist charismatic people the public already respects. In climate he names four wildly different figures who shifted opinion and even business rhetoric in recent years: Pope Francis, David Attenborough, Bill Gates, and Greta Thunberg. Fridman notes that Gates and Thunberg also create division; charisma without trust is not enough. Another social-media cost: among young people Rees knows, fewer will contemplate a political career because of the pressure on them and their families. The quality of those who choose the path is dropping, and we all lose. Fridman still believes a hunger for leaders will pull the brave in. He adds Musk to the climate list. Rees is a fan of the engineering—someone has called Musk a 21st-century Brunel—and is unsure he is an ethical inspiration.

Advice for the young, offered diffidently. Realize the choices you make now matter; Rees has watched friends miss open doors because they thought those chances belonged only to a privileged few. In science, pick a field where new things are happening, so you can work on questions the old guard never had a chance to pose. Do not enter a stagnant subject and chew on the problems they already got stuck on. Then pick a style you can stand: solitary thought, fieldwork, a big team, computing, mathematical derivation. Stay flexible. Keep your eyes open. Most people, himself included, did not see this until too late. Fridman says it applies well beyond science.

Is he afraid of death? No. He feels lucky to have lasted this long and to have had a fortunate life compared with most people. The current craze for living much longer—Altos Labs in San Francisco, La Jolla, and Cambridge, funded by billionaires who wanted to be rich when young and now want to be young again—will not be so easy. Do we want it? An elite that lives far longer than everyone else would be a fundamental inequality. If it happened to everyone, that might be an improvement, and it is not obvious. For his part, living as long as most people, with luck, is all he can expect and a lot to be grateful for. Fridman closes with Rees’s own words: most educated people know we are the outcome of nearly four billion years of Darwinian selection, and many still think humans are the culmination. The Sun is less than halfway through. It will not be humans who watch it die six billion years from now. Any creatures then will be as different from us as we are from bacteria or amoebae.