In the United States, a university professor and his students have developed the first chip based on machine learning to optimize storage and playback on hard drives. This AI allows for more capacity and lower latency when accessing data.
With more and more data, mechanical hard drives remain the true workhorses of high-tech. More storage capacity is always needed, especially in the cloud. Whether it is to use the apps of a smartphone or any connected gadget, or to manage what the sensors collect in a city or in the industry, nothing can work without a real-time reading of this data, and an Artificial Intelligence finds itself milling in the void.
The concern is that a hard drive is ultimately quite rudimentary. The data is piled up there anyhow and they come to be stored in a sparse way a little everywhere on the memory. Despite fast hard drives, the time to access this sometimes essential data remains problematic, especially when this data is stored in the cloud.
So the obvious idea is to increase performance by optimizing storage. Easy to say, but to do it is more complicated. In any case, this is the challenge launched by a researcher from Carnegie Mellon University in Pennsylvania, USA, with his students.
Reduce latency, increase storage capacity
To achieve this, they developed a chip to analyze the data to directly optimize their storage. The technology is based on an Artificial Intelligence that learns on its own how to better manage information to optimize its storage and reading. On an ordinary hard disk, the data recorded on the magnetic layer is identified via signal processing technology. With the reinforcement of AI, the storage process is optimized, with less loss of storage space, and performance suffers.
In the end, even in the cloud, the latency for calling data can be reduced with this process. This chip, based on a neural network, should arrive before the end of the year in the form of a physical prototype in the lab of the university. This potentially revolutionary development is closely watched by hard drive manufacturers who even say they are surprised that a university laboratory has been able to create such a prototype.