In 1889, a French doctor named Francois-Gilbert Viault descended from a mountain in the Andes, took blood from his arm and examined it under a microscope. The red blood cells of Dr. Viault, which carries oxygen, increased by 42 percent. He discovered a magical power of the human body: When it needs more of these vital cells, it can make them on demand.
In the early 1900s, scientists believed that a hormone was the cause. They called the theoretical hormone erythropoietin, or “red maker” in Greek. Seven decades later, researchers found the actual erythropoietin after filtering 670 gallons of urine.
And about 50 years after that, biologists in Israel announced that they had found a rare kidney cell that produces a hormone when oxygen drops too low. It’s called the Norn cellnamed after the Norse gods believed to control human destiny.
It took people 134 years to discover the Norn cells. Last summer, computers in California discovered them on their own in just six weeks.
The discovery occurred when Stanford researchers programmed computers to teach themselves biology. The computers ran an artificial intelligence program similar to ChatGPT, the famous bot that became fluent in the language after training on billions of pieces of text from the internet. But the Stanford researchers trained their computers on raw data about millions of real cells and their chemical and genetic makeup.
The researchers did not tell the computers what these measurements meant. They did not explain that different types of cells have different biochemical profiles. They don’t specify which cells detect light in our eyes, for example, or which make antibodies.
Computers crunch the data themselves, creating a model of all the cells based on their similarity to each other in a vast, multidimensional space. When the machines are done, they have learned an amazing amount. They can classify a cell they have never seen before as one of over 1,000 different types. One of those is the Norn cell.
“That’s remarkable, because nobody said in the model that a Norn cell exists in the rock,” said Jure Leskovec, a computer scientist at Stanford who trained the computers.
The software is one of several new AI-powered programs, known as foundational models, that set their sights on the fundamentals of biology. Models don’t just organize the information that biologists collect. They make discoveries about how genes work and how cells develop.
As the models grow, with more laboratory data and computing power, scientists predict they will begin to make deeper discoveries. They can reveal secrets about cancer and other diseases. They can figure out the recipes for turning one type of cell into another.
“An important discovery about biology that biologists would otherwise not have made — I think we’ll see that at some point,” said Dr. Eric Topol, the director of the Scripps Research Translational Institute.
How far they will go is a matter of debate. While some skeptics think the models will hit a wall, more optimistic scientists believe the foundational models will even address the biggest biological question of them all: What separates life from the inanimate?
