Brain Cells in a Dish Learn How to Play Pong
Scientists have taught a collection of brain cells living in a bowl how to play a version of the arcade game Pong. The research could one day provide doctors with a “sandbox” to test treatments for brain diseases.
For hundreds of years, the scientific community has attempted to unravel the inner workings of the human brain. This hyper-complex organ contains around 86 billion specialized messenger cells — known as neurons — that control everything from how we mediate our vital bodily functions to how we conjure up and express complex thoughts.
Unlocking the mysteries of its function would allow scientists to cure countless diseases and advance a range of related technologies.
To this end, some of the brightest minds on earth have created myriad computer models of the brain at varying scales and levels of complexity. However, an international team of scientists is trying a different approach by taking embryonic mouse brain cells and human brain cells made from stem cells and growing them on a microelectrode array.
This array is able to track the behavior of the 800,000 cells and apply electrical stimulation to trigger activity in them. In fact, DishBrain, as the team calls it, is a relatively simplified living model of part of a living brain.
“In the past, models of the brain have been developed aimed at how computer scientists think the brain might work,” comments Dr. Brett Kagan, lead author of the new study and chief scientific officer at Cortical Labs. “That’s usually based on our current understanding of information technology, such as silicon computing. But the truth is, we don’t really understand how the brain works.”
In a new study published in the journal Neuron, scientists took DishBrain and tried to get the cells to act in an intelligent and coordinated way to get a task done. More specifically, they wanted to see if they could get the myriad cells to act as a unit and successfully play the tennis game of pong.
The team used a series of electrodes to create their virtual pong court. Using electrical signals, they were able to tell the cells which side of the pitch the ball was on, and the frequency of those signals was used to indicate its direction and how far the ball was from passing through an invisible wall to score to achieve.
According to a press release from Australia’s Science in Public, the feedback from the electrodes was also used to teach the model brain how to return the ball. More specifically, the activity of cells in two defined regions of the dish was captured and used to move a virtual paddle up and down.
However, training the model brain to move the paddle correctly was a challenge. Normally, dopamine is released by the brain to reward correct action, and this in turn encourages a subject to act in a certain way. With DishBrain, that wasn’t an option.
Instead, the team turned to a scientific theory known as the “free energy principle,” which states that cells, like neurons, will do whatever it takes to reduce the unpredictability in their environment.
The team implemented the theory by hitting the bowl with an unpredictable electrical stimulus when the paddle failed to intercept the ball, after which the virtual ball would take off again on a random vector. Conversely, when the neurons were able to move the racquet to successfully deflect the ball, a predictable electrical stimulus was applied to all cells simultaneously, after which play continued in a predictable manner.
As the cells tended to make their surroundings predictable, they worked to understand the game and prolong the pong rally.
“The beautiful and groundbreaking aspect of this work lies in providing the neurons with sensations – the feedback – and above all the ability to act on their world,” says Professor Karl Friston, co-author of the new study from University College London. “Remarkably, cultures have learned to make their world more predictable by reacting to it.”
The team discovered that DishBrain’s ability to extend a rally improved significantly over the course of just five minutes. In other words, the cells were able to organize themselves to achieve a goal using what the researchers defined as synthetic biological intelligence.
‘The translational potential of this work is really exciting: it means we don’t have to worry about creating ‘digital twins’ to test therapeutic interventions,’ comments Professor Friston. “We now have what is essentially the ultimate biomimetic ‘sandbox’ in which to test the effects of drugs and genetic variants – a sandbox made up of exactly the same (neural) computational elements found in your brain and mine.”
In the future, researchers plan to give DishBrain alcohol to see how it affects his performance at pong. The authors of the study hope that the model could one day provide a viable alternative to animal testing and provide physicians with new insights into degenerative diseases such as dementia.
Anthony Wood is a freelance science writer for IGN
Photo credit: Cortical Labs