How many brain cells do you need to play a video game? At first glance, this question may seem like a deviant or enigmatic or even humorous question; But the fact is that today scientists have found a scientific answer to it. Thanks to the neural network system called DishBrain, there is now a real answer to the said question. According to new research, to do Pong game About 800 thousand brain cells will be needed.
Although the one-sided and slow strategy of these cells for digital pong will naturally not be able to win championships in electronic sports or similar competitions in the near future, their performance shows the capacity of combining living tissues with silicon technology for more successful applications in the future.
The recent experiment is the first test of artificial biological intelligence and shows that neurons can adjust their activity to perform a specific task. More importantly, neurons can learn to do that work and do it better in the future if they are given the right feedback.
The recent achievement is very surprising and will have potentially valuable applications in the field of computing and studying a variety of brain-related phenomena, from the effect of drugs and treatments on brain activity to how intelligence develops in the first place. Brett Kagana neuroscientist from a biotechnology startup Cortical Labs In Australia it says:
We have shown that we can interact with living biological neurons in such a way that they can be induced to change their activity and [این کار درادامه] lead to the emergence of something similar to intelligence.
The brain is a combination of nerve cells; Neurons derived from embryonic mice and human neurons grown from stem cells. These cells were grown on arrays of microelectrodes. The mentioned arrays could be activated in order to stimulate the neurons and, in other words, provide the sensory input needed for research.
Microscopic image of nerve cells. Neurons, axons, and dendrites glow purple, red, and green under the microscope with fluorescent markers.
The microelectrodes located on both sides of the game frame showed the position of the ball on the left or right side of the paddle in order to play pong. This work was done while the frequency of the signals simultaneously transmitted the distance of the ball and its position.
Dishbrain can move the paddle to hit the ball only by relying on these settings; However, at least at this point in general, it has a very poor performance. Neurons need feedback to play good games. The research team in this field developed a software to provide feedback through electrodes every time the ball is lost.
This allowed the system to achieve improvements and improve performance in the pong game. Researchers’ observations indicate that the process of learning and improving performance lasts only 5 minutes. Carl FristonA theoretical neuroscientist from University College London in the United Kingdom says:
The beautiful and groundbreaking aspect of this work is equipping neurons with elements related to sensations and feedback, and more fundamentally, the ability of neurons to act based on the world around them. Human cultures have learned to make their world more predictable, and this is important; Because we cannot do this kind of self-organization [به سیستم فعلی] to train, only because (unlike a pet) these little artificial brains have no sense of reward or punishment.
A few years ago, Freestone proposed a theory called The principle of free energy brought up According to this theory, all biological systems behave in such a way as to reduce the gap between what is expected and what is experienced. In other words, they try to make the world more predictable for themselves. According to Freestone, the dishbrain is simply doing what biology does best by adjusting its actions to make the world more predictable. Kagan says:
We chose the pong game because of its simplicity and familiarity; But this game was one of the first games used in machine learning. So, we wanted to find out how it works.
An unpredictable stimulus is applied to the cells, and subsequently, the system as a whole reorganizes its activity to better play the game and minimize random response. Also, it can be considered that the mere act of playing and hitting the ball and receiving predictable stimulation will in itself create somewhat more predictable environments.
Such a system has really interesting capabilities; Especially in artificial intelligence and computing. The human brain contains about 80 to 100 billion neurons and is far more powerful than any other computer, and our best computers strive to replicate or mimic its performance. Our best effort to date also required 82,944 processors and one petabyte of main memory and 40 minutes of time to replicate just one second of the activity of 1% of the human brain.
If the architecture of artificial systems becomes more similar to the structure of the real brain, achieving better performance records will not be far away; Perhaps even an artificial biological system like the one Kagan et al. developed could handle such a task.
In addition, we should probably pay attention to other results of these experiments. Consequences and outputs that are likely to have faster results compared to the main purpose of the research. For example, Dishbrain may be able to help chemists better understand the effects of different drugs on the brain at the cellular level. It is even possible that one day, using neurons cultured from stem cells reverse-engineered from the patient’s skin, we will be able to produce drugs tailored to the patient’s specific biology.
According to Friston, the translational potential of this work is really exciting. From this evaluation, it follows that we should not worry about creating “digital twins” in order to conduct trials of therapeutic interventions. In fact, we now have the ultimate biomimetic “sandbox” in which we can test the effects of drugs and all kinds of genetic variables.
The interesting and valuable thing about the sandbox described by Freestone is that it is composed of exactly the same computing (neural) elements found in our human brains. Now, the next step is to understand how Dishbrain’s ability to play pong is affected by drugs and alcohol. Kagan says:
Currently, we are trying to create a dose response curve with ethanol; So basically put them under the influence of alcohol and see if they play the game less well! Just like when alcohol affects people.
Apart from everything, it would be very interesting to evaluate the response of a number of neurons to familiar stimuli. The result of this team’s research Published in the journal Neuron.