Scientists has successfully Uploaded a Brain in to a computer
A fruit fly is walking right now — not in a jar on a laboratory bench, but inside a computer.
Scientists have taken a real fly’s brain, mapped every single neuron and connection, and run it inside a simulated body. The result is a digital fly that behaves like a living one, responding to its digital environment entirely on its own.
It represents the first time a complete biological brain has been digitally copied and made to control a physical body, opening the door to deeper understanding of how brains work, the discovery of new intelligence algorithms, and even the distant possibility of human brain emulation.
Understanding the Fly Brain
Mapping of the Fruit Fly Brain
The brain used in this experiment belongs to the common fruit fly, Drosophila melanogaster. Despite its tiny size, the fly’s brain has:
- ~125,000 neurons
- ~50 million synaptic connections
Scientists mapped these neurons using high-resolution electron microscopy. The data was assembled through the FlyWire project, which has carefully reconstructed the entire neural networks in three dimensions.
Every neuron was catalogued, including:
- which neurons connect to which
- how strong those connections are
- whether the neurons excite or inhibit other neurons
This full wiring diagram of the brain is known as a connectome. You can think of it like the blueprint of a building — except instead of walls and pipes, it contains neurons and synapses.
From Neurons to Action
Once the connectome was mapped, researchers turned it into a running simulation.
Each neuron in the system was given a simplified mathematical model called a leaky integrate-and-fire neuron. This model determines when a neuron activates based on incoming signals. Importantly, the system contains:
- excitatory neurons that encourage activity
- inhibitory neurons that suppress activity
- connection strengths determined by the number of synapses between neurons
Notably, this model does not include learning or memory. The simulated fly cannot form new long-term memories. Its behaviour comes entirely from the existing structure of its brain.
Giving the Brain a Body
The Sensorimotor Loop
A brain alone cannot produce behaviour unless it can interact with a body and environment.
To achieve this, researchers connected the simulated brain to a realistic fly body built using NeuroMechFly, running inside MuJoCo physics engine software.
This created a complete sensorimotor loop:
- The simulated environment sends sensory input to the brain.
- Neural activity propagates through all 125,000 neurons.
- Motor commands are generated.
- The simulated body moves.
- Those movements generate new sensory input.
This loop repeats continuously, allowing the fly to interact with its environment in real time.
The result is a fly that this walks, turns, and respond to stimuli — all driven directly by its neural wiring.
Researchers estimate the behaviour of the simulated fly matches real flies with about 91% behavioural accuracy.
Why This Is a Breakthrough
Scientists have attempted whole-brain simulations before, but those efforts typically fell into two categories:
Brains without bodies
The neural activity was simulated, but it had no effect on the physical world.
Bodies controlled by AI
The body moved realistically, but the control system was trained with machine learning rather than built from real neural wiring.
For example, the OpenWorm project simulated the worm Caenorhabditis elegans, which has only 302 neurons.
In contrast, the fruit fly brain contains more than 400 times as many neurons.
What makes this experiment unique is that it combines:
- a full biological connectome
- a functioning neural simulation
- a physics-based body
- a responsive environment
Together, these elements produce multiple natural behaviours driven purely by the brain’s circuitry.
Current Limitations
Despite the breakthrough, the system still has several limitations.
The researchers could not directly map every motor neuron controlling the fly’s body, because the body itself was not scanned during the connectome reconstruction. Instead, they had to connect known movement signals to the simulated body.
The neuron model also lacks plasticity, meaning the simulated brain cannot learn or adapt.
Also, the environment in which the fly operates is still relatively simple. The research team intends to expand it over time to create richer and more natural environments.
From Flies to Larger Brains
The ultimate goal is to scale this approach to larger animals.
For perspective, here are approximate neuron counts across species:
| Organism | Neurons |
|---|---|
| C. elegans | 302 |
| Fruit fly | ~125,000 |
| Mouse | ~70 million |
| Human | ~86 billion |
The next major target for researchers is the mouse brain, which is roughly 560 times larger than the fruit fly’s brain.
Mapping and simulating such a brain will require enormous advances in imaging, data storage, and computational power.
However, the fly experiment demonstrates that the core concept works.
What This Could Enable
Whole-brain emulation could eventually unlock several important capabilities.
Understanding neurological diseases
Scientists could simulate disorders such as Parkinson’s or Alzheimer’s and test treatments directly on virtual brains.
Discovering new intelligence algorithms
Evolution has been refining biological intelligence for hundreds of millions of years. Simulating these systems may reveal computational strategies that artificial intelligence has not yet discovered.
Digital brain emulation
In theory, if every neuron and synapse in a human brain could be mapped and simulated with sufficient accuracy, the resulting system might reproduce the behaviour and cognition of that person. However, the neurons in the human brain are much smaller, so we are currently unable to accurately map very well.
Ethical Questions
The research team has acknowledged the ethical implications of this work. Because scientists cannot yet determine whether simulated neural systems have subjective experiences, they take the possibility seriously. As a result, the team intends to provide richer environments for simulated organisms rather than confining them to minimal test setups. This approach reflects a growing discussion about how digital biological systems should be treated as the technology advances.
The Bigger Picture
For decades, the phrase “the ghost in the machine” has been used to describe consciousness. In this experiment, something unusual is happening: the machine itself is beginning to reproduce the structure of the ghost.
Just a few years ago, whole-brain simulations were limited to tiny worms with a few hundred neurons. Now scientists are running a system with over 125,000 neurons controlling a body in real time. The pace of progress is accelerating.
Each step brings us closer to a future where biological intelligence can be mapped, simulated, and explored inside digital environments — potentially transforming neuroscience, artificial intelligence, and our understanding of the mind itself.
Even though these brain simulations are producing very similar behaviours, its unlikely these simulations are conscious.
References
-
Wissner-Gross, A. (2026) — The First Multi-Behavior Brain Upload. Eon Systems blog post describing the first embodiment of a connectome-based fruit fly brain controlling a simulated body.
-
Shiu, P. et al. (2024) — Whole-brain computational model of the adult Drosophila brain. Published in Nature. Introduced a computational model of ~125,000 neurons and ~50 million synapses derived from the fruit fly connectome.
-
FlyWire project (Princeton University and collaborators) — High-resolution reconstruction of the Drosophila melanogaster brain connectome used as the basis for the neural simulation.
-
NeuroMechFly — A biomechanical simulation framework that models the fruit fly body and locomotion.
-
MuJoCo physics engine — Physics simulation platform used to run the embodied fly simulation.
-
Andregg, M. (2026) — Public explanation thread on X describing the architecture and behaviour accuracy of the uploaded fruit fly brain simulation.
-
OpenWorm — Earlier whole-organism simulation project modelling the nervous system of Caenorhabditis elegans, containing 302 neurons.
-
Turaga, S. et al. / Janelia Research Campus — Large-scale connectomics research enabling reconstruction of insect neural circuits used in FlyWire datasets.
