In 2022, engineers at the Israeli company Cognifiber announced that they were able to create a new photonic chip. This chip is planned to be used for peripheral computing, which eliminates the need to connect to cloud structures. The company plans to do everything possible so that over time, photonics will be able to replace silicon, which is the main component in all modern chips. Glass, on the other hand, is a much more affordable and cheaper material, which will help solve many problems. For example, it will turn out to reduce the cost of training artificial intelligence several times.
At the same time, Cognifiber is not just engaged in photonic computing. Here, photonics is given special attention. Engineers use it as a communication center between racks, develop innovative microcircuits that can independently perform calculations. All this in the future will be used by both small companies and large data centers.
Separately, it is worth highlighting the words of Zeev Zalewski, who is the technical director of the company. In his opinion, this decision is the future of the development of artificial intelligence. The use of photonic glass in chips will bring the learning process to peripheral devices. Now it is difficult to do this due to capacity and power limitations.
The technologies used by the company’s engineers are based on fiber optic computing. Its essence lies in the use of analog networks of machine learning. Previously, this technique was used to process data in a fiber optic cable. Now, engineers have begun to use this approach on processors with a glass substrate. Cognifiber believes that their chips can be programmed in the same way as conventional chips.
There is every reason to believe their claims. The matter is that already now the system supports linear and non-linear operations. Thanks to this, it can be used to launch artificial intelligence. With the help of this technology, in the future it will be possible to increase the performance of computers by tens or even hundreds of times. In addition, this will affect the cost of training artificial intelligence.
Thanks to the use of photonic chips based on glass, as well as special fiber optic cables, it will be possible to reduce the size and bring powerful servers to the periphery. Such a solution will allow to make a real breakthrough in machine learning and artificial intelligence training.
Cognifiber CEO Eyal Cohen believes that their technology will be in demand in all areas of activity where large amounts of data are processed. Apparently, this is actually the case. Indeed, in 2022, the company had no problems with financing. Already, Cognifiber is finalizing the technology. And if everything goes without changes, then in 2023 the commercial production of photonic chips will begin.