Alexander Graham Bell posed outdoors with teachers and students of a school for deaf children, 1883

A vizmaya story about judgment

Would AI have told Alexander Graham Bell no?

Three famous scientists. Three answers an AI would have trusted. One skill that matters more than writing good prompts.

vizmaya desk · 23 October 2026

Photo: Bell with teachers and students, 1883. Library of Congress, Gilbert H. Grosvenor Collection, via Wikimedia Commons, public domain.

First principles

An AI answers from what people have already written.

It reads a huge amount of past writing. Then it predicts a good answer. That makes it excellent on old questions.

A brand new question has nothing written about it yet.

1875

A teacher of deaf children asks a strange question.

Alexander Graham Bell was 29. He taught deaf students. He was a teacher, not an electrician.

His question: can a machine carry my voice to someone far away?

Bell outdoors with teachers and students of the Scott Circle School for deaf children in Washington, D.C. — Bell with teachers and students of the Scott Circle School, 1883. Library of Congress via Wikimedia Commons, public domain.

The record in 1876

The telegraph was a giant. Bell had one rough device.

By 1876, Western Union ran over 7,000 offices and 185,000 miles of wire. It was worth about $55 million.

Late that year, Bell's team offered it every telephone patent for $100,000. Its president, William Orton, said no.

Source: American Heritage, "Hindsight, Foresight, and No Sight"

Imagine Bell with an AI in 1875

A well-read AI would have said no.

Imagined reply. Not a real transcript.

Short answer: no. Speech is a very weak signal. The best electrical minds are building telegraphs that send dots and dashes.

That is a fair answer. It matches everything written in 1875.

The twist

It would also have fixed Bell's mistake. That would have hurt.

What Bell believed

A German physicist, Hermann von Helmholtz, had sent vowel sounds down a wire.

What was true

Helmholtz made vowel sounds with tuning forks and electromagnets. Nothing travelled far on a wire.

Bell could not read German. He studied the diagrams and misread them. That wrong belief gave him the nerve to start. A perfect reader would have removed it.

Historians say this story is often told too neatly. Treat it as a strong push, not the only one. Source: ATL Society, Gorman, University of Virginia

He tested it anyway

Patent to first words took three days.

Bell's lawyer filed the patent on February 14, 1876. A rival, Elisha Gray, filed a warning notice the same day.

The patent was issued on March 7. On March 10, Bell's assistant, Thomas Watson, heard Bell's voice through the machine.

Source: US National Archives

Page 2 of Bell's 1876 telephone patent drawing, showing the device and its parts — Bell's telephone patent drawing, page 2, 1876. US Patent Office, National Archives, via Wikimedia Commons, public domain.

Once it was real

The "no" described 1875. The telephone lived in 1876.

700%

By one account, how far Bell company stock rose in 17 days, after Western Union agreed in November 1879 to leave the telephone business.

Source: Schiavo, "The Bell-Western Union Agreement of 1879"

1898, Paris

Marie Curie: one number would not go away.

Atoms were thought to be solid and unchanging. Curie measured how strongly materials give off rays, as uranium does.

Pitchblende, a uranium ore, gave about 4 times the signal of pure uranium oxide. Uranium alone could not explain that.

Source: Marie Curie in Lapham's Quarterly

Portrait photograph of Marie Curie around 1898 — Marie Curie, about 1898. Author unknown, via Wikimedia Commons, public domain.

Imagine Curie with an AI

An AI would have blamed the equipment.

Imagined reply. Not a real transcript.

Most likely a measurement problem. Uranium explains the signal. Re-check your instruments.

Good advice. She re-checked. The number stayed. So she reasoned that the ore must hide something new.

Four years of work

Then she made proof you could hold.

900x

radium's signal compared with uranium, December 1898

0.1g

of pure radium chloride by 1902, from several tons of ore

Her first new element, polonium, was named for her home country, Poland.

Sources: IUPAC, Cambridge blog

Photograph of Marie Curie around 1920 — Marie Curie, about 1920, via Wikimedia Commons, public domain.

1905, Switzerland

Einstein: for three years, the best data said he was wrong.

Albert Einstein lecturing in Vienna in 1921 — Einstein lecturing, Vienna, 1921. Photo: Ferdinand Schmutzer, via Wikimedia Commons, public domain.

Most physicists thought light moved through an invisible "ether". Einstein dropped it. New electron data then seemed to favor a rival idea. Better data in 1908 backed Einstein.

The pattern

Each was right. Each needed three to four years to show it.

Until then, the written record said no. A model trained on that record would have said no too.

Bell: 1876 to the 1879 settlement. Curie: 1898 to isolated radium in 1902. Einstein: 1905 to better data in 1908. Source: Wikipedia, Schiavo, Lynge

Being fair to AI

Use AI for what is written. Test what is not.

Strong at

Reading a paper in German
Finding who filed first
Checking the arithmetic

Blind at

Questions nobody has answered
Odd results that break the pattern
Which of two sources is broken

A no from an AI costs seconds. A real test costs months. Use the seconds to decide what to test.

The skill

Four questions before you trust a no.

1
Is the answer already written down?
Then use AI, and check one source.
2
Is anyone doing this for the first time?
Then a no only reports the past.
3
Can I test it cheaply?
Ask: Give me the cheapest experiment that could prove you wrong.
4
Is the no a law of nature or a habit?
Ask: Label each reason: physical law, engineering limit, or habit.

Sources & method

How we checked this

The AI replies in this story are imagined. No scientist used one. Two widely shared Bell stories could not be confirmed and are left out: a Western Union memo calling the phone "idiotic", and a $25 million buyback offer. More vizmaya stories.

Photos: Wikimedia Commons. Bell and students: Library of Congress. Patent drawing: US National Archives. Curie: author unknown. Einstein: Ferdinand Schmutzer. All public domain.