Trending
Time flies when I'm thinking about flies, and more specifically about the fruit fly, Drosophila melanogaster. This weekend, I entertained myself by asking questions and writing jokes. I wondered whether there are other animals named after the actions they take. A fly flies, I thought, but an elephant doesn't elephant. ChatGPT, an artificial intelligence tool, informed me that ducks duck, golden retrievers retrieve hunted prey, and woodpeckers peck.
As for jokes, here is one:
Fly 1: "Why does everyone think I'm a dangerous criminal?"
Fly 2: "I don't know. Why?"
Fly 1: "Because every time I show up, they bring out the SWAT team!"
My kids told me the joke was consistent with my other "daddy jokes." I thought it was funny. Gemini, another artificial intelligence tool, sided with me, as it often does. It's very agreeable.
AI can easily generate jokes that rely on wordplay. But can it come up with the kind of humor professional comedians create? People like Jerry Seinfeld find humor in ordinary human situations. Their jokes rely on sharp observations about how people think and behave. AI can imitate but not duplicate this kind of humor.
Why am I thinking about fly puns and jokes? Why am I so excited about flies all of a sudden?
Because something remarkable has happened. In 2024, scientists published a map of the entire brain of an adult fruit fly, including its visual-processing regions. The brain is smaller than a poppy seed, yet it contains about 140,000 neurons--nerve cells--connected through more than 50 million synapses, the tiny junctions where neurons communicate.
The achievement required electron microscopy, enormous computing power, artificial intelligence, and years of human work. The result--a detailed map showing which neurons connect to which--is called a connectome.
But a map just sits there. The next question was irresistible: Can you make it do something? In March 2026, a San Francisco startup called Eon Systems took a model based on the published connectome and connected it to a virtual fly body. Information from the simulated environment entered the system, activity traveled through the modeled neural circuitry, and the output drove the fly’s movement.
Then things became even stranger. Once the fly's neural wiring became publicly available, independent programmers began experimenting with it. Connectome-based models were hooked up to video games, including Minecraft, Super Mario 64 and Doom.
To be clear, the fly's brain did not suddenly become a tiny gamer sitting in front of Doom, plotting how to kill demons.
The experiments were much cruder than that. Images or information from a game were converted into inputs that a simulated neural network based on the fly's wiring could process. Activity traveled through the modeled circuitry, and selected outputs were translated into commands: move, turn, jump, or perform some other action.
For simplicity, I think of it somewhat like a self-driving car. The car receives information about roads, traffic lights, pedestrians, and other vehicles. That information is processed, and the system produces an output: turn right, turn left, accelerate, slow down, stop.
But there is a fascinating difference. The architecture of a self-driving system is engineered and trained by humans. The architecture of the fly's brain comes from biology. It is the product of evolution, development and experience--a tiny information-processing machine refined over hundreds of millions of years.
Human brains, too, are not blank slates. We are born with biological machinery shaped by evolution and development, machinery that allows us to learn to recognize faces, walk, speak, think, and navigate a complicated social world.
Which made me wonder: If we can map a fly's brain and use its architecture to control a virtual body--or even stumble through a video game--what happens if one day we can do something similar with a human brain?
The scale is almost absurd. The human brain contains roughly 86 billion neurons and perhaps 100 trillion synapses. Mapping all those connections would be a monumental undertaking. And even a perfect wiring diagram would not necessarily recreate a mind. Brains are more than wiring: neurons change, connections strengthen and weaken, chemicals alter their behavior, and experience continuously reshapes the system.
Still, imagine what we might learn.
Instead of designing artificial intelligence entirely from scratch, perhaps we could borrow more from the solutions evolution has already discovered. Could biological architecture help us build machines capable of different kinds of reasoning or creativity? Could it help us solve problems our current AI systems cannot? Could it someday compose more inspiring music or tell funnier jokes?
And if it can, what will be left for us to do?
Perhaps we can still swat flies.
EDITOR'S NOTE: Shahar Madjar, MD, MBA, is a urologist and an author. He practices at Schoolcraft Memorial Hospital in Manistique and in Baraga County Memorial Hospital in L'Anse. Find his books on Amazon or contact him at smadjar@yahoo.com.