It was to be expected. There is so much talk about artificial intelligence that our next PCs will soon all be equipped with a chip dedicated to AI. This way the main processor will be less overloaded and this will cause fewer slowdowns.
Published
Update
Reading time: 2 min
A dedicated AI chip is what we call an NPU, a neural processing unit. Basically, it’s a microprocessor that accelerates everything related to artificial intelligence. Currently, it is the main processor of the machine which takes care of this. However, as it is not designed for that, we sometimes observe slowness.
To avoid this problem, most applications (image or text generators like ChatGPT) actually work remotely, on supercomputers connected to the internet, which simply return the result to you. In the future, by integrating these types of chips into PCs, we should see more and more applications running locally.
Apple was the first to integrate a dedicated chip into its Macs with M1, M2 and M3 processors. Intel and AMD followed. The first PCs are starting to be marketed. And eventually, all machines should be equipped with it.
A system already in place on mobile phones
On mobile phones, we don’t necessarily know, but this has already been the case since 2017! Since the launch of Samsung’s Galaxy S8 and Apple’s iPhone Moreover, it is the arrival of these dedicated chips which made it possible to switch to unlocking by facial recognition. It is also thanks to them that phones can do things that PCs cannot yet, such as copying and pasting text from an image, or adding subtitles, in real time, to n any video.
All this is done today, with the snap of a finger, locally, on telephones. Whereas on a PC, even a much more powerful one, you have to go through internet services, or see the computer “grind”.
Initially, these computers with “special AI” chips will cost more. These dedicated chips arrive first on high-end PCs. Then they will gradually be integrated into all machines, as was done with telephones. This will not be an option. This will be offered as standard, for a whole new generation of AI-powered applications.