Plate Nº 72 · recorded November 17, 2026

Neuroscience & MindReported finding

Fruit fly brain map inspired Doom and bitcoin experiments

A public map of more than 166,000 fruit-fly neurons inspired Doom, bitcoin-trading and Rubik’s Cube experiments, but scientists stress that no living fly’s mind was uploaded.

By James Calloway4 min read773 words

In brief

  1. Early September: scientists unveiled the first in-depth connectome of a male fruit fly’s brain and nervous system.
  2. The atlas covers more than 166,000 neurons, including their shapes, sizes and connections.
  3. It is the fourth extensive fly connectome and the first built from a male fly.
  4. The dataset builds on two decades of brain research and may support comparisons between male and female brains.
  5. Programmers used it for Doom and bitcoin-trading sites; Nick Walton also trained it to solve a Rubik’s Cube.
Virtual fruit fly brains can play Doom and trade bitcoin
Plate Nº 72Virtual fruit fly brains can play Doom and trade bitcoin — Kost9n4 / pixabay

By James Calloway

More than 166,000 neurons were mapped in a male fruit-fly connectome released in early September, and programmers soon used the public dataset for Doom and bitcoin-trading experiments.

What did scientists map?

A connectome is a detailed map of nerve cells, including their shapes, sizes and links to one another. Researchers unveiled the first in-depth map of a male fruit fly’s brain and nervous system in early September, prompting celebration among neuroscience researchers.

  • The atlas describes more than 166,000 neurons and the connections among them.
  • It is the fourth extensive connectome of its kind in flies and the first one made from a male specimen.
  • It builds on two decades of research aimed at understanding how brains work.
  • Researchers consider the male map particularly useful for examining similarities and differences between the sexes.

Fruit flies are a key animal model because studying their brains can illuminate pathways and functions that humans share. The new dataset gives researchers a detailed public resource for that work, rather than evidence that artificial intelligence has acquired a fly’s consciousness.

What did programmers actually do?

Public access changed the project from a specialized neuroscience resource into an experimental playground. People set the virtual “fly” to work in environments ranging from the video game Doom to bitcoin-trading sites, without reporting measures of game performance or financial returns.

Gerald M. Rubin, head of biology at the Howard Hughes Medical Institute in Virginia and one of the researchers behind the dataset, described the reuse as unexpected but fun. He said it spotlights the research and gives the general public an entry into an otherwise complicated scientific topic.

Bing Wen Brunton, a computational neuroscientist at the University of Washington in Seattle, said many programmers loaded the connectome data into an artificial-intelligence model. They then trained that software on a defined task, such as playing chess or video games.

The model’s parameters are adjustable numerical settings that guide its responses. They change whenever the virtual fly gets closer to completing the task, Brunton said, which causes the represented connectome to shift during training.

Brunton summarized the process this way: “It’s being modified, sometimes by a lot, until it accomplishes the task the programmer has set out for it to do.”

Did a fly really play Doom?

No. Brunton described real neurons as processing machines: they receive inputs, use different parameters to determine how to respond, and often release signals to connected neurons. The reported shifts occurred in adjustable software settings, not in a living fly’s anatomy or behavior.

Brunton studies how neural activity relates to the physical body. She said researchers remain far from a science-fiction-style brain upload into a virtual environment or machine, and people should not assume the flies themselves played the games or traded bitcoin.

She nevertheless defended the public interest. If the broader public engages with the dataset and has fun, Brunton asked, what else do scientists want?

What made the project spread online?

Nick Walton, founder and CEO of the AI game company Latitude, had played with the dataset himself. Among his experiments, he trained it to solve a Rubik’s Cube.

Walton noted that scientists had made connectome datasets public before. This release coincided with the rise of AI “vibe coding,” while its popularity grew from the “memeability” of flies doing absurd things in videos.

Walton knows no fly’s mind exists inside his computer, but he remains interested in the overlap between biological intelligence and artificial intelligence. He said: “Is this way more advanced than just training something with a machine-learning model? Not really,” adding, “It’s more a capture of this underlying idea of what happens if you upload beings with biological intelligence into a digital world.”

His formulation describes a metaphor built around a neural map, not an actual upload of a living creature.

What does this change for people who saw the demos?

It changes how to interpret the viral experiments. Viewers saw software trained with a detailed biological wiring diagram, not sentient insects operating Doom or making investment decisions. The project supplies no reported evidence of consciousness, human-like understanding or profitable bitcoin trading.

The practical change is access. Developers now have an unusually detailed public dataset for building and testing artificial-intelligence systems, while neuroscientists gain another male-brain map for studying shared brain functions and possible sex-based differences.

The atlas is a substantial research resource. The playful demos did not create a digital fly mind. They showed how far a public dataset can travel when programmers combine biological structure with machine learning.

via scientificamerican.com (Source)

Filed under

  • fruit-fly
  • connectome
  • brain-mapping
  • artificial-intelligence
  • machine-learning
Share this article:

More from James Calloway

James Calloway

Show full bio

Staff writer covering regulation, policy and their impact on general science news at SciBeat.

474 articles

Nearby plates

« Previous articleNext article »