Google Maps Fruit Fly Brain with 166,000 Neurons, Engineers Immediately Train It to Play Doom

Reviewed byNidhi Govil

7 Sources

Share

Google and HHMI Janelia Research completed the first complete brain map of an adult male fruit fly, containing 166,000 neurons and 125 million synaptic connections. Within days, software engineers trained the simulated fly brain to play Doom, Super Mario 64, and other games, turning a neuroscience breakthrough into an unexpected development testbed.

Google Maps Entire Brain of Adult Male Fruit Fly

Google Research, in collaboration with HHMI Janelia Research Campus and other scientific partners, announced a decade-long breakthrough in neuroscience: the complete mapping of an adult male fruit fly brain and central nervous system

1

5

. The mapped fruit fly brain, released as the MaleCNS v1.0 dataset, contains over 166,000 neurons and 125 million synaptic connections, making it the largest brain map by neuron count to date

3

4

. This AI-powered 3D model represents a foundational resource for neuroscience, promising to accelerate understanding of how brains perceive the world and react to stimuli.

Source: Tom's Hardware

Source: Tom's Hardware

The research process involved taking extremely thin slices of a fly's brain and body, imaging them with electron microscopes, and using AI-enabled brain mapping to combine millions of 2D images into 3D neural shapes

1

. Scientists then spent years verifying and annotating the neural connections to create what neuroscientists call a connectome—essentially a wiring diagram showing how neurons are connected

5

. Google emphasized that mapping the roughly 86 billion neurons in a human brain remains impossible with current technology, making fruit flies an ideal starting point for understanding neural computation.

Fly Brain Plays Doom Within Days of Release

Just days after Google released the MaleCNS v1.0 dataset, software engineers began training the simulated fly brain to play classic video games. Coinbase software engineer Alex Wormuth created a system where each Doom frame generates 3,335 brightness inputs and 811 R8 color inputs that stimulate sensory neurons in the simulated biological neural networks

1

2

. These inputs hit specific neurons based on what the frame shows, and the resulting neural activity is mapped to game controls. When the player takes damage, the system triggers a stimulus to two PPL101 dopamine cells—the same receptors that control hunger and aggression in real fruit flies—as reinforcement

1

4

.

Wormuth made the code fully open source and set up a live page showing the training process, asking "Will the fly learn to survive?"

3

4

. His Doomfly website displays both the fly's point of view and a third-person perspective, allowing visitors to watch sensory input, neural activity, and other metrics in real time as the simulated fly brain learns to shoot demons

1

.

Source: TweakTown

Source: TweakTown

Simulated Fly Brain Tackles Super Mario 64 and Beyond

Developer Jessica Paquette quickly followed with a similar experiment, training the fruit fly brain to play Super Mario 64

1

3

. Her video showed Mario repeatedly jumping and bumping into walls, suggesting the system remains in early training stages

4

. Paquette described the project as "literally 100% vibe coded with GPT Astra... just for fun" and released the code on GitHub

3

5

.

The experiments expanded rapidly across the developer community. Mechanistic interpretability researcher Lyra Bubbles trained the fly brain to play Beat Saber, while other developers created fly brain projects for escape rooms, soccer games, and even Minecraft

1

5

. Wormuth later gave the simulated fly brain $100 worth of bitcoin to trade, and another developer made it deploy its own cryptocurrency token on Robinhood

1

. One X user fine-tuned the system to make Y-M-C-A poses in response to different tones, while another showed it the viral Bad Apple!! animation to observe its reactions

5

.

Why This Neural Connectome Matters for AI and Neuroscience

The immediate transformation of the MaleCNS v1.0 dataset from scientific resource to public development testbed demonstrates how quickly biological data can inform computational experiments

4

. Instead of building artificial neural networks from scratch, developers now have access to a detailed biological neural network they can test in software. This approach offers insights into how known biological connections behave when exposed to input and feedback, potentially bridging neuroscience and artificial intelligence research.

Source: PC Gamer

Source: PC Gamer

For neuroscience, the connectome provides a foundation for studying how neural circuits influence behavior and how damaged neural pathways might eventually be repaired

5

. Fruit flies have served genetics and neuroscience research for over a century, and this complete nervous system map extends that legacy into the AI era

3

. Google is already working with researchers on connectomes for fish and mice, suggesting larger biological models will become available for both scientific study and computational experimentation

5

. Watch for more fly brain projects to emerge as developers continue exploring what biological neural computation can teach us about intelligence, learning, and adaptation.

Today's Top Stories

© 2026 TheOutpost.AI All rights reserved