Connect with us

Hi, what are you looking for?


Groundbreaking Research Develops Neural Network Using Real Neurons Resulting in Impressive Output

Image Source: sdecoret / Shutterstock

Neural networks have gained immense popularity in recent years, being utilized for a variety of tasks like image recognition, text generation, and gaming. These artificial neural networks are essentially complex mathematical models running on computers, which, while powerful, have not yet demonstrated true intelligence.

Cortical Labs, a research facility based in Melbourne, Australia, has taken a unique approach by growing actual biological neurons on electrode arrays to interface them with digital systems. Their latest research, detailed in a pre-print paper awaiting peer review, suggests that these biological neural networks have the potential to learn and adapt.

Exploring Neuroscience

The primary objective of this research is to leverage the computational capabilities of biological neurons to create “synthetic biological intelligence.” Unlike computer-simulated neural networks, biological neurons offer significantly more complexity and functionality. Thus, the rationale is to utilize biological neurons rather than rely solely on simulations to develop truly intelligent systems.

The research team experimented with neural networks grown from mouse and human cells. Mouse cortical cells were sourced from embryos, while human pluripotent stem cells were differentiated into cortical neurons for testing. These cells were cultured on a high-density multielectrode array provided by Maxwell Biosystems.

Once the cells were cultured and settled onto the electrode array, they formed intricate interconnections across the surface. The cells could then be electronically stimulated via the electrodes, with the neural responses being monitored. The resulting system was dubbed DishBrain, symbolizing neural matter living in a controlled laboratory environment.

DishBrain underwent testing in a simulated game environment similar to Pong. The biological neural network (BNN) was equipped with electrodes that provided sensory input based on the game state, with additional electrodes controlling the paddle’s movement in the game.

Various feedback mechanisms were employed to train the neural network to maneuver the game intelligently. The Free Energy Principle, a concept where biological systems strive to maintain alignment between external stimuli and internal models, guided the design of the feedback loops. Different feedback conditions were tested, including Stimulus, Silent, No-Feedback, and Rest modes, with the Stimulus mode showing learning capabilities over time.

Advertisement. Scroll to continue reading.

The study’s preliminary results indicated that the neural network under the Stimulus condition had longer rallies and improved gameplay performance compared to the other modes. Notably, the network’s performance enhanced over time exclusively under the Stimulus condition, indicating a learning effect. Conversely, the performance under the Silent and No-Feedback modes remained relatively constant throughout the test.

While the results are yet to undergo peer review, they hold significant promise. The successful interfacing and training of neuronal cells not only advance our understanding of brain function but also pave the way for enhancing artificial intelligence through biological components. The research hints at the possibility of growing and training simple biological brains to perform cognitive tasks with digital interfaces.

The documentation, while intricate, highlights the potential of training biological neurons to intelligently cooperate with digital systems. While it’s early in the research stage, there’s immense potential for integrating biological brains with technology, possibly revolutionizing various fields in the future.

Image Source: sdecoret / Shutterstock

You May Also Like


In the past, updating the Linux kernel to include new features or support for additional hardware was a common practice, with users needing to...


Ever been on the verge of the ultimate techie setup, only to stumble upon the need for one last tweak? For Maxime, the final...


Recently, we explored the complexities surrounding timezones and the Moon’s unique timing system. While Earth’s time is Sun-based, the Moon follows a different rhythm...


In the past, high-voltage power supplies were common in households for tasks like electron acceleration and generating X-rays for television, but with the advancement...