The term AI has taken the world by storm in the past few months. Appliances, cars, even health care devices all claim to have artificial intelligence models. But many of these AI labels are a bit dubious. This confusion extends to computers, with the advent of the AI PC. What really makes a computer an AI PC?

First, let’s examine what an AI PC really does. With regard to Windows, current AI PC models are built around Microsoft’s Copilot+. Many users have become familiar with Copilot, Microsoft’s cloud-based “intelligent” assistant, over the past few months. Copilot+ is a bit different. It provides features such as the somewhat controversial Recall, which runs locally on the computer, not in the cloud.

Recall basically takes snapshots of what you’ve worked on, allowing you to easily remember what you’ve done in the past and pick up where you left off. But this could be a big problem if someone gets unauthorized access to your computer.

PC manufacturers add their own features to AI PCs as well. Many vendors promise improved productivity. This is accomplished by tracking what you do with your computer, when you do it, and optimizing the system to work best for those tasks. Other improvements, common among virtually all AI PCs, include easier and faster content creation as well as improved collaboration, which provides smarter, dynamic adjustments that allow you to look and sound better in online meetings.

With regard to Macs, Apple doesn’t have Copilot+ (though there is an app available that is widely panned). The Mac AI is called Apple Intelligence. Of course, content creation has always been their sweet spot and Intelligence promises improvements there.

What about ChatGPT? For the foreseeable future, ChatGPT and similar large language models, better known as LLM, will continue to run exclusively in the cloud, not locally on an AI PC. It just takes too much processing power. That’s not to say that you can’t run ChatGPT on your AI PC. But you won’t have any advantage running any cloud-based AI services, including LLM, on an AI PC, over just about any other reasonably powered computer.

Don’t miss out on what’s happening!

As to technical specs, there is a primary factor differentiating AI PC’s: the presence of a Neural Processing Unit, referred to as NPU. An NPU takes AI-specific workloads off the other processors in the computer, namely the main Central Processing Unit, or CPU, and the Graphics Processing Unit, or GPU (woefully misnamed as it has been handling more than graphics for years now).

NPUs don’t need a lot of power and are specifically designed for complex computing. As such, they are highly efficient at contemporary AI-related tasks such as facial recognition, voice assistance and image processing.

Determining whether a computer has an NPU takes a bit of investigation. Currently, all NPU are packaged with the same chip as the CPU. Take a look at the kind of CPU (often generically called the processor) your computer has to see whether it includes an NPU.

Of course, venerable Windows processor manufacturers Intel and AMD have lines of chips that include NPU. Examples include Intel’s Core Ultra and AMD’s Ryzen AI.

A more recent entry in the PC processor space is Qualcomm. Their Snapdragon X lines are prevalent in AI-labeled computers from all the big boys, including Lenovo, Dell, HP and Microsoft.

Apple’s NPU, which they call Neural Engine, is part of their M-series line. All reasonably current Macs have an NPU; the M1 was released almost 5 years ago.

But of course, buyer beware. Once you identify the exact model of CPU in a computer, a simple online check will tell you if it has an NPU. Since new features for AI PCs continue to be introduced, if you’re in the market for a new computer it would be prudent to ensure you get a model with an NPU.

John Agsalud is an information technology expert with more than 25 years of IT experience in Hawaii and around the world. He can be reached at jagsalud@live.com.