A team of neurotechnology researchers at the company Eon Systems has achieved something that until recently seemed impossible: digitally reconstructing the entire brain of a fly and running it within a virtual environment that controls its behaviour autonomously. In other words, there is no external programming. This experiment marks a turning point in the convergence of neuroscience, artificial intelligence, computer simulation and cognitive science, and probably also in our understanding of what consciousness really means.
From the connectome to emergent behaviour
To understand the magnitude of the breakthrough, it is important to understand the concept of the connectome - the connectome is the complete map of all the neurons and their synaptic connections within a brain. In the case of the fruit fly (Drosophila melanogaster), this map contains approximately..:
- ≈ 140,000 neurons
- ≈ 50 million synapses
This neural map was previously reconstructed by international scientific projects using electron microscopy and advanced data analysis techniques. The Eon Systems team used this connectome as the basis for creating a neuron-by-neuron computational model using leaky integrate-and-fire (LIF) simulations, a widely used approach for modelling neural dynamics. The digital brain was then connected to a virtual body within a physical simulation.
The result was surprising: the digital fly began to move, react to stimuli and execute basic behaviours, such as walking and scouring the environment for food, without having been previously trained by machine learning algorithms. In other words, the behaviour emerged directly from the neural architecture.
A different paradigm from today's AI
This experiment raises a fundamental question about the future of artificial intelligence. Today, the dominant paradigm in AI is based on training models with large volumes of data, such as language models or deep learning systems.
But the approach presented here suggests an alternative: Instead of training an intelligence, we could reconstruct it from its biological architecture. If this approach proves scalable, it could completely redefine the roadmap to artificial general intelligence (AGI).
The challenge of climbing
While the experiment is remarkable, it is also important to contextualise its limits. The current model does not reproduce all aspects of the biological brain, for example:
- Complex neurochemical processes
- Interaction with glial cells
- Adaptive learning mechanisms
- Complete neuronal plasticity
Still, it represents a critical proof of concept: it demonstrates that brain structure alone can generate coherent behaviour within a simulated environment. The real challenge will be to scale this approach, to put it into perspective:
- Fly → ~140,000 neurons
- Mouse → ~70 million
- Human beings → ~86 billion
Complexity increases exponentially. However, the history of technology has taught us that many advances begin as seemingly small demonstrations.
Scientific, technological and philosophical implications
This raises an uncomfortable question:
If we can completely replicate a brain... what are we really replicating?
What exactly have we created?
- A computational model
- A digital agency
- Or something much more complex and profound?
Is this a new paradigm for building artificial intelligence, or perhaps the first step towards creating conscious digital organisms?
References:
https://eon.systems/updates/embodied-brain-emulation
Author:
David Hernan Tardini
Ph.D: Engineering Project Management
Expert in technological innovation and corporate agility