In 1889, a French doctor named François-Gilbert Viaud descended from a mountain in the Andes, took blood from his arm, and examined it under a microscope. Dr. Viaud's red blood cells, which carry oxygen, increased by 42 percent. He had discovered the mysterious power of the human body. That means if we need more of these vital cells, we can make them on demand.
In the early 1900s, scientists theorized that hormones were to blame. They called the theoretical hormone erythropoietin, Greek for “red maker.” Seventy years later, researchers discovered actual erythropoietin after filtering 670 gallons of urine.
And almost 50 years later, Israeli biologists announced that they had discovered a rare kidney cell that produces a hormone when oxygen gets too low. It's called a Norn cell, named after the Norse gods who were believed to control human destiny.
It took 134 years for humans to discover Norn cells. Last summer, California's computer discovered them automatically in just six weeks.
The discovery came when researchers at Stanford University programmed a computer to help people teach themselves biology. The computer was running an artificial intelligence program similar to ChatGPT, a popular bot that learned to speak the language fluently after training it with billions of texts on the internet. But researchers at Stanford University trained their computer using raw data about millions of real cells and their chemical and genetic makeup.
The researchers did not tell the computer what these measurements meant. They did not account for the fact that different types of cells have different biochemical profiles. They did not define, for example, which cells capture light in our eyes or which cells make antibodies.
The computer processed the data on its own and created a model of every cell based on their similarities to each other in a vast multidimensional space. When the machine was completed, they had learned an amazing amount. They were able to classify him as one of over 1,000 different types of cells he had never seen before. One of them was a Norn cell.
“This is surprising because no one had ever taught the model that Norn cells were present in the kidney,” said Jure Leskovec, a computer scientist at Stanford University who trained the computer.
The software is one of several new AI-powered programs, known as fundamental models, that are targeting the fundamentals of biology. Models do more than simply organize the information biologists are collecting. They are making discoveries about how genes work and how cells develop.
As models scale up and laboratory data and computing power increase further, scientists predict they will begin to make more profound discoveries. They may reveal secrets about cancer and other diseases. They might find a recipe to turn one type of cell into another.
“I think we're going to see important discoveries in biology made by biologists that otherwise wouldn't be made by biologists.” Scripps Translational Institute Director Dr. Eric Topol said.
How far they will go is a matter of debate. Some skeptics think the model will hit a wall, but more optimistic scientists believe the basic model will solve the biggest biological problem of them all: what separates life from non-life. I believe it will even work on what it is.