WILL COMPUTER CHIPS EVER RESEMBLE THE MIND?
Scientists out of the University of Munster in Germany seem to think so. They have recently succeeded in creating a piece of hardware that they believe will pave the way for creating computers resembling the human brain. They have created a chip that contains a network of artificial neurons that use light to communicate with one another. This is designed to resemble the way the neurons in our brains communicate with one another via electrical signals traveling through synapses. This chip is capable of learning and evaluating very large quantities of data.
We have mentioned before that when neural networks work together they can engage in what is referred to as deep learning in the world of AI. The idea that these neural networks could one day emulate the human brain continues to drive us towards these kinds of innovations. Given the information that we have it is safe to say that eventually a cluster of neural networks will resemble our minds. How capable they will be or whether they will actually be conscious or not is a different story.
This graphic illustrates the basic idea behind neural networks. The fact that we don’t have a great understanding of what goes on inside the hidden layer is the basis of the black box problem covered in a previous post.
LIGHT IS MUCH FASTER THAN ELECTRONS
The researchers demonstrated that this kind of neural network is able to actually “learn” and use what it learns to recognize patterns in a nearly identical way to our own minds. As stated before their system works with light instead of traditional electrons. This means it can process data much faster. It is a photonic system that uses light to send signals from one “neuron” to the next.
There is vast potential here for evaluating patterns in larger data sets than was previously possible. There is also far more potential for learning than there ever was with previous neural networks.
HOW DOES IT WORK
Neural Networks have been around since the 80’s. The approach back then was a neuromorphic system based on electronics. Newer optical systems in which photons are used instead are brand new. These optical systems work with optical wave-guides that can transit light and can be fabricated into microchips which are integrated with phase change materials. These materials are already found on storage media such as DVDs. The phase change materials change their optical properties based on how the atoms arrange themselves.
They can be arranged in a crystalline fashion which is a regular fashion. Or they can be arranged in an irregular fashion which is referred to as amorphous. In this irregular fashion the phase change can be initiated by light.
The lead author of the experiments that led to this light-based computer chip says “Because the material reacts so strongly, and changes its property dramatically, it is highly suitable for imitating synapses and the transfer of impulses between two neurons.”
Basically, when the atoms arrange themselves in response to heat or any adequate stimuli information can be transferred from one photon to another at the speed of light. This is an improvement upon the technology that allows DVDs to work the way that they do.
THE POTENTIAL OF THIS TECHNOLOGY
We have already established that by working with photons instead of electrons data can be transferred extremely fast, which opens the floodgates for advanced neural network based AI programs that can accomplish far more than ever before. The chip created at the University of Munster also has an edge over traditional neural networks in storing and processing information.
A practical example of its capabilities lie in the detection of cancer cells. Using the light-based computer chip one could identify cancer cells automatically and faster than with other methods. They only limitation is the number of neurons and synapses which make up the overall depth of the network. The next step is to take this light-based chip and expand it as far and as deep as possible to see what it is capable of.