The interaction between the human brain and machines is becoming more and more similar. Newly developed memory resistors are a candidate to take an important step in this field.
The progress of traditional computers is now starting to slow down. During this period, different information technologies had an opportunity to make progress. One of them is computers compatible with the way the human brain works.
A team of researchers from Penn State University took an important step towards transmitting the neural network structure in the human brain and the analogue nature of our brains to computers. Machines will thus be able to communicate better with the human brain.
Conversion from digital to analog
Our modern computers are digital and work with 1s and 0s. There are many more situations in which analog systems like our brains can be found. Simply put, the computer turns the light on or off. Our brain can adjust the brightness to the desired level.
According to Saptarshi Das, leader of the research team at Penn State University, computers inspired by the brain and called Neuromorphic have actually been the subject of studies for 40 years. Today, areas such as fast visual processing and seeing patterns in big data stand out as areas where systems similar to our brain are more successful and are in great need.
As Das stated, the problem in digital systems stems from the fact that data is processed and stored in different places. This causes a serious loss of energy and speed. It also greatly increases the required storage space.
New neural network model
Saying that they create artificial neural networks and want to reflect energy and space savings in the brain with these systems, researcher Thomas Shranghamer explains the power of the brain with an interesting example. Saying that we carry our brains above our heads, the researcher states that supercomputers take up as much space as a few tennis courts.
In the newly developed method, data and processing capacity are loaded into a graphene layer. The use of this data and process capacity is shaped according to which region and how intensity the electricity supplied in the process is carried out. Thus, multiple memory states can be kept on a single graphene surface.