Microsoft Copilot will run locally on PCs

Microsoft is preparing the ground to ensure a significant evolution in the use of artificial intelligence on PCs with the Copilot service. This is because it will run from the device. This means that it will leave aside its dependence on the cloud, according to Intel revealed to the Tom’s Hardware website.

read more

According to the chip-making company, Microsoft plans to operate its AI chatbot locally “soon,” although it has not yet announced a specific timeline. This would be to facilitate small tasks on the platform, which is often overloaded due to the high volume of simultaneous accesses and requests.

This means that not all AI will operate directly on the device. Many of the more complex content creation functions will continue to require cloud services. This suggests that issues such as latency, performance and privacy could be improved with some components being processed directly on the device.

This will result in necessary adaptations to the technical specifications of the devices. Intel had already announced that, for a chipset suitable for AI, the neural processing unit (NPU) must be capable of executing at least 40 TOPS, that is, 40 trillion operations per second.

The latest chips from manufacturers like Intel, AMD and Apple don’t currently reach 20 TOPS. This indicates that there is still a long way to go to meet the performance demands set by Microsoft.

Furthermore, for local execution of AI functionalities, devices must have at least 16 GB of RAM. If the computers do not meet these criteria, Copilot and other chatbots will continue to work via the cloud, but without the previously mentioned benefits.

In addition to being integrated into the operating system, Microsoft announced that it will introduce a dedicated Copilot hotkey on keyboards, as well as the ability for instant activation in the system.

Source: Connected World