AI Is About to Break Settled Science

TLDR: Every generation has some closely held scientific belief that has ultimately proven to be very wrong. Below are ten times the smartest people alive defended a belief against the data, and lost. I read these examples as a preview of the next decade, when AI enabled by a new generation of instruments start measuring things that were never measurable before, and lead to scientific insights that disrupt some of our core beliefs.

When I was studying molecular genetics and aerospace engineering at MIT in the early 1980s, the number of confirmed planets outside our solar system was zero. Today NASA’s count is past 6,000, and the Kepler data says the Milky Way holds more planets than stars (and the NASA Roman Space Telescope will increase these numbers somewhere between 1,000X and 1,000,000 fold). Nothing about the galaxy changed. Our instruments did.

That pattern runs through the entire history of science. The evidence usually shows up before the belief changes. A telescope sees something it shouldn’t. A flask refuses to spoil. A microwave antenna picks up a hiss that everyone wants to write off as noise (aka the afterglow of the Big Bang). Then comes the hard part, which is never the measurement. It is changing our minds.

I’ve said for years that the day before something is a breakthrough, it’s a crazy idea.

The evidence usually shows up before the belief changes. Changing the belief is the hard part.

Here are ten crazy ideas that turned out to be reality, and what each one cost us in cherished assumptions.

1. EARTH IS NOT THE CENTER OF ANYTHING

Stand outside. The ground is still, the Sun moves, the stars wheel overhead. An Earth-centered cosmos was the obvious reading of the data, and it anchored astronomy for roughly 1,400 years after Ptolemy.

Copernicus published the heliocentric model in 1543. In 1610, Galileo aimed a telescope at Jupiter and found four moons orbiting it. Things demonstrably revolved around something other than Earth. Venus showed phases that Ptolemy’s model could not produce. Kepler and Newton finished the job over the following century.

The psychological cost was enormous. Humanity had to give up its cosmic address at the center of it all. Our experience had been real; our interpretation of it had been wrong. Keep that distinction in mind. It comes up nine more times.

2. THE EARTH IS NOT 6,000 YEARS OLD

Biblical chronologies put creation about 6,000 years back, a timeline you could hold in your head. In 1788, James Hutton took John Playfair and James Hall by boat to Siccar Point on the Scottish coast and showed them tilted ancient rock lying beneath younger horizontal layers. For that to exist, sediment had to accumulate, harden, tilt, erode, and get buried all over again. Playfair later wrote that “the mind seemed to grow giddy by looking so far into the abyss of time.”

Hutton never calculated an age. Radiometric dating eventually did: 4.54 billion years, per the USGS. All of human civilization occupies the final sliver of that record. The cliff had not changed in 1788. Our ability to read it had.

3. DISEASE COMES FROM MICROBES, NOT BAD AIR

For most of medical history, epidemics were blamed on “miasma,” foul air. In the 1860s Louis Pasteur boiled broth in swan-neck flasks that let air in but trapped particles; the broth stayed sterile. On March 24, 1882, Robert Koch announced the bacterium that causes tuberculosis, a disease killing roughly one in seven people in Europe and the Americas at the time.

An invisible cause became something you could isolate, culture, and kill. Antiseptic surgery, vaccines, antibiotics, and clean water all follow from that one shift. The smallpox eradication campaign alone is estimated to save on the order of 5 million lives every year since 1980. As a physician, I’d put germ theory at the top of this list for sheer human impact.

4. SPECIES ARE NOT FIXED

Each species was assumed to be separately created and permanent. Darwin and Wallace independently arrived at natural selection; their papers were read together at the Linnean Society on July 1, 1858, and On the Origin of Species followed in 1859. Individuals vary, some variations improve reproductive success, and those differences accumulate across generations.

Humanity got a family tree that extends into every living thing. Today evolution organizes all of biology, and the sequencing revolution lets us read that family tree letter by letter. The first human genome cost $3 billion in 2003. Today it costs under $300 dollars.

5. THERE IS NO UNIVERSAL CLOCK

A second was a second everywhere. Einstein’s special relativity (1905) and general relativity (1915) said otherwise: motion slows elapsed time and mass curves spacetime. The 1919 eclipse expeditions measured starlight bending around the Sun by the predicted amount.

This one is now engineering rather than philosophy. GPS satellite clocks run about 38 microseconds a day fast relative to the ground because of relativity; without the correction, your phone’s position would drift by roughly 10 kilometers per day. An assumption can feel fundamental simply because our senses never encounter the conditions where it breaks.

6. REALITY IS PROBABILISTIC AT ITS CORE

Classical physics promised that with perfect knowledge of starting conditions you could predict everything. Atoms refused to cooperate; their light spectra made no classical sense. Heisenberg, Born, Schrödinger and others built quantum mechanics in the mid-1920s, and in 1927 Heisenberg showed that position and momentum cannot both be sharply defined. The limit is in the math, not in the equipment.

Every transistor, laser, and MRI machine on Earth runs on this theory. Nature was speaking a different language than the one we had assumed, and once we learned it, we got the entire digital economy as a side effect.

7. THE UNIVERSE IS EXPANDING AND HAD A BEGINNING

Space was assumed to be a fixed stage. In 1927 Georges Lemaître derived expansion from Einstein’s equations; in 1929 Edwin Hubble showed that distant galaxies recede faster the farther away they are. Einstein had inserted a fudge factor into his own equations to keep the universe static, and reportedly called it the biggest blunder of his career.

Then in 1965 Penzias and Wilson at Bell Labs detected the cosmic microwave background, relic heat from a universe roughly 13.8 billion years old. The stage itself has a history, and our instruments can read its first chapters.

8. THE CONTINENTS MOVE

In 1912 Alfred Wegener noticed matching coastlines, fossils, and rock formations on opposite sides of the Atlantic and proposed continental drift. He had no mechanism, and geologists dismissed him for 50 years. In the 1960s, ocean-floor mapping and seafloor magnetic stripes revealed plates spreading at roughly the speed your fingernails grow, 2 to 5 centimeters a year.

Wegener’s mechanism was wrong and his observation was right. That distinction matters for anyone evaluating a new idea today. You can detect something real long before you can explain it.

9. DNA, NOT PROTEIN, CARRIES THE CODE

Biochemists bet on proteins to carry heredity. Proteins were complex; DNA looked chemically boring with only four bases. In 1944 Oswald Avery, Colin MacLeod, and Maclyn McCarty showed purified DNA could permanently transform bacteria. In 1952 Alfred Hershey and Martha Chase used radioactive labels to show that viral DNA, not protein, entered the infected cell. Watson and Crick published the double helix the next year.

Our intuition about what complexity should look like misled us. Four letters turned out to be enough to write every organism that has ever lived, and today we edit those letters with CRISPR in a graduate student’s afternoon.

10. PLANETS ARE EVERYWHERE

In 1995 Michel Mayor and Didier Queloz found 51 Pegasi b, a Jupiter-mass planet orbiting its star every four days, an arrangement our own solar system said was impossible. The space telescope Kepler then established that planets outnumber stars in the Milky Way, and a 2020 analysis of Kepler data estimated roughly 300 million potentially habitable Earth-sized worlds in our galaxy alone.

The worlds were always there. They were beyond our instruments’ reach. Better measurement turned a possibility into a population.

WHAT WE ARE WRONG ABOUT RIGHT NOW

Every entry on this list follows the same arc: a new instrument, a measurement that doesn’t fit, a decade or a century of resistance, then a new reality. The instruments arriving this decade are unlike anything that came before. AlphaFold predicted structures for essentially all 200 million known proteins; human crystallographers had managed about 200,000 in the previous 50 years. The Vera Rubin Observatory is imaging the entire southern sky every few nights. Single-cell sequencing, continuous molecular monitoring, and whole-brain imaging are generating data faster than any human can read it, and AI is now reading it for us.

So here is the fork. You can spend the next decade defending the map, or trusting the instrument. History has not been kind to the first group.

My own candidates for the next demolition: that aging is inevitable rather than a treatable process, that intelligence requires a biological brain, and that Earth is the only place in this galaxy with life. I’d bet at least one of those falls within 20 years.

“You can spend the next decade defending the map, or trusting the instrument. History has not been kind to the first group.”

WHAT THIS MEANS FOR YOU

If you’re an entrepreneur: The biggest companies of the next decade will be built on measurements that were impossible five years ago. Pick a field where a new instrument just arrived and ask what everyone still believes that the data no longer supports.

If you’re an executive: Wegener was right for 50 years before anyone acted on it. Audit which “settled” assumptions in your industry are simply old, and decide in advance what data would change your mind.

If you’re an investor: Every entry above looked like fringe science until the year it didn’t. Put a small slice of capital into teams making previously unmeasurable things measurable.

If you’re a student: Textbooks lag the data by a decade. Learn to read the instruments directly. The skill that compounds is knowing what evidence would change your mind, then going to find it.

If you’re a parent: Teach your kids that “everyone knows” is a claim, not a fact. Ten of the smartest generations in history were certain and wrong. Theirs will be too, and the ones who notice first will lead.

Pick one belief you consider settled. Write down what evidence would change your mind. Then ask whether an instrument that exists today, or an AI reading data nobody has bothered to look at, could produce it.

The next time science rewrites reality, you’ll want to be the one holding the telescope.

To a future of abundance,

Peter