Machine learning and artificial intelligence are seen as a serious tool to influence the entire industry. Many applications for self-propelled photography, Netflix suggestions, and digital assistants are using this technology, and in the future these technologies will become even more important. That’s why Intel is beginning to produce AI-compatible chips to take advantage of this technology.
Almost everything that can be defined as neural networks, machine learning algorithms, and artificial intelligence is now using graphics cards. Most of the learning and activities are done by GPUs. However, GPUs can handle this data advantageously, but they are not built to perform these operations, so they are far from ideal processors when AI is the case. GPU may be much weaker than a special AI card, as a CPU with general purpose has poorer graphics processing than a GPU.
Intel believes that with Nervana’s new platform and chip, it can quadruple performance. With this chip built from the ground up for AI applications, Intel says it can be completed in less than 100 times less than teaching a deep learning model.
The new chip class, Lake Crest, will begin testing in the first half of 2017, which will be commercially available sometime later. According to Intel vice president Diane Bryant’s statement, it is expected that the Intel Nervana platform will cause a serious performance increase and a dramatic decrease in the training time of neural networks.
However, although Intel says that Nervana is the first chip designed for AI, it is worth noting that this is not exactly true. Google has also produced a silicon chip called the Tensor Processing Unit, and IBM is working on a commercial product that will be released in the future.
Currently there is not much information about the full talents of Nervana and Intel’s future enthusiasts. However, if the incredible speed of development in this area and the incredible potential of AI are considered, it will be easy to hear more about it in the near future.