Saturday, March 2

Mouse neurons grown on a lab plate learn to play the video game Pong

The creation of a synthetic biological intelligence seemed, until now, the stuff of science fiction. A team of Australian researchers has taken a decisive step in that direction by teaching mouse brain cells placed in a kind of laboratory dish to learn to play Pong. a classic and simple table tennis video game marketed by the Atari company in 1972. The results of the study are published today in the magazine neuron.

A rapper built with Artificial Intelligence is sacked for reproducing racist stereotypes

Know more

The researchers announced their results last December in a preprint in the magazine bioRixv, but it is now that his experiment has been independently collated and validated by scientists. To ‘teach’ the cells to play, the researchers have used a support that they have named DishBrain (something like a ‘brain dish’). It’s actually an electronic version of the so-called Petri dish or dish: the shallow, round containers used in laboratories to grow cell cultures.

For this project they have used between 800,000 and one million cells from the cerebral cortex of rodents cultured together with human stem cells called ‘pluripotent’ (which, when developed, can give rise to very different tissues).

Scientists have long been using neurons in artificial devices containing multiple microelectrodes and reading their activity. Neural signals are sent and obtained through these devices (technically known as microelectrode arrays) –electric currents of ions–, which is why they are used as connections between neurons and electronic circuits.

We know that our brains have the evolutionary advantage of having been fine-tuned over hundreds of millions of years for survival. Now we have the possibility to take advantage of this incredibly powerful and cheap biological intelligence

Dr Adeel Razi
Director of the Systems and Computational Neuroscience Laboratory at Monash University.

The novelty now is that, for the first time, the researchers have managed to stimulate brain cells in a “structured and significant” way – they indicate in their study – so that they carry out a task aimed at a specific objective. In this case, play Pong.

“We have shown that we can interact with living biological neurons in such a way that we force them to modify their activity, which leads to something that resembles intelligence,” says lead author of the project, Brett Kagan, who is scientific director of the biotech company Cortical Labs.

The DishBrain would then be a kind of mini braina hardware biological, what is known by the anglicism wetwarealluding to the wet character –wet, in English– of the biological substances used. One of the objectives of the project is to study the possibilities of taking advantage of biological computing, which is much more powerful than artificial computing. In fact, those mini brains created by Australian scientists learned to play Pong much faster than a common artificial intelligence.

“We know that our brains have the evolutionary advantage of having been fine-tuned over hundreds of millions of years for survival. Now, it seems that the possibility of harnessing this incredibly powerful and cheap biological intelligence is within our grasp,” says Dr Adeel Razi, Director of the Systems and Computational Neuroscience Laboratory at Monash University.

Another of the objectives of the research is to be able to study the effect of substances such as alcohol, to find out exactly how they affect neurons. “We tried to create a response curve to different doses of ethyl alcohol: basically getting the cells drunk and seeing if they play worse, the same thing that happens when people drink,” says Kagan. The findings also raise the possibility of creating an alternative to animal testing when investigating how new drugs or gene therapies respond in these dynamic settings.

The paddles and the ball

To carry out the experiment, the team used mouse cells from embryonic brains, as well as some human brain cells derived from stem cells, and grew them on arrays of microelectrodes that can be stimulated and whose activity is readable.

Electrodes to the left or right of an array fired to tell DishBrain which side the ball was on, while how far away the paddle was was indicated by the frequency of the signals. The feedback from the electrodes taught DishBrain how to return the ball, making the cells act as if they were the paddle themselves.

“Never before have we been able to see how cells act in a virtual environment,” says Kagan in a press release released by Cortical Labs: “We managed to build a closed-loop environment that can read what is happening in cells, stimulate them with meaningful information and then change the cells interactively so that they can actually be altered.”