A way of turning every system into a neuron computer node that will cause a revolution in the field of artificial intelligence

The idea as a whole is known as 'repository computing' and originates from attempts to develop computer networks that are a model of the brain. The idea is based on the assumption that it is possible to decode the behavior of any physical system - from a bucket of water to drops of plastic wrapped in carbon nanotubes - with the aim of utilizing the computing power inherent in them

[Translation by Dr. Nachmani Moshe]
Researchers claim that any system that undergoes changes can be used as an efficient computer.

The latest chip on which the iPhone 7 is based includes 3.3 billion transistors packed inside a piece of silicon the size of a small coin. However, the trend of ever smaller and more powerful computers may be coming to an end. Silicon-based chips are reaching a point where the laws of physics prevent them from continuing to shrink. In addition, there are several important limitations in the performance of chips of this type. Therefore, many researchers around the world are looking for other alternatives to silicon chips.

The most well-known alternative is quantum computers, which utilize the features of the chips in a completely different way than normal digital machines. In addition, there is the possibility of using alternative materials - potentially any material or physical system - capable of performing calculations, without the need to control the electrons, as is the case with silicon chips. The idea as a whole is known as 'repository computing' and originates from attempts to develop computer networks that are a model of the brain. The idea is based on the assumption that it is possible to decipher the behavior of any physical system - from a bucket of water to drops of plastic wrapped in carbon nanotubes - with the aim of utilizing the computing power inherent in them.

Repository computers utilize the physical properties of the material in its natural state in order to perform part of the calculation. This is in contrast to the current digital computing model where the features of the computer are changed in order to perform the calculation. For example, in order to create modern microchips we change the crystalline structure of silicon. A storage computer could be composed, in principle, of a piece of silicon (or of other materials) without the need to make these changes to the material itself. The basic idea is to excite the material in a certain way and then study and measure how the excitation affects it. If we can learn how to get from the input (the excitation) to the output (the change), then we are essentially performing a calculation that we can then use as part of a set of calculations. Unlike conventional computer chips that rely on the position of the electrons, the particular arrangement of the particles within the material is not important. Instead, we just need to observe certain general properties that allow us to measure the change in matter.

For example, a research team built a simple reservoir computer from a bucket of water. The researchers demonstrated how, after exciting the water with mechanical tests, they were able to teach a camera aimed at the surface of the water to read the distinct wave patterns that were created. Then, they matched the calculation associated with the test movements to the wave pattern, and used the result to execute a number of simple logic commands. In principle, the water itself transferred the input from the test into a useful output - and therein lies the important insight.

It turns out that the idea of ​​a storage computer aligns well with recent research in the field of neuroscience in which researchers discovered that parts of the brain are intended for 'general use'. These areas are mainly composed of collections of neurons whose arrangement is quite loose and yet they are able to support cognitive functions that occur in more organized areas of the brain, thus making the whole process more efficient. As happens in a computer, if a pool of these neurons is excited by a defined signal, it responds in a very characteristic way, and this response can help perform calculations. For example, recent research suggests that when we hear or see something, a certain general part of the brain is stimulated by the corresponding signal - sound or light. The response of the neurons in this area of ​​the brain is then transmitted to another area of ​​the brain, which is of higher specialization.

The research indicates that storage computers could be extremely robust and powerful, and in addition, in theory, could perform an infinite number of functions. In fact, stimulus repositories have already become especially popular in certain areas of artificial intelligence, thanks to these very properties. For example, systems that use database methods to predict trends in the stock market and the economy have been proven to be much more effective than conventional artificial intelligence technologies. Ultimately, this is still a relatively new technology, and further research is required as to the capabilities and implications of this method. At the same time, it is already clear that there is a huge number of possible applications for this type of technology, both in the field of artificial intelligence and in a variety of other fields, from real-time data analysis and processing to image/pattern recognition and robot control.

The original article