AI is a hot topic in technology right now, so it’s not entirely surprising to see a whole range of PCs – desktops and laptops – come out with AI branding on them. It’s important to recognise what AI means in the personal computer space – and what it can and can’t do.
What does AI mean for personal computing?
When you hear about AI in computing, it’s most likely going to be based around what’s typically called “Generative AI”. That’s the AI where you put in a prompt and it creates something for you based on that prompt, whether it’s a text file, an image, even a piece of music. Generative AI systems get built up around analysis of thousands of pieces of content so that they can generate results that are meant to be in line with your expectations, and while they’re getting smarter and better over time, there’s still room for improvement.
You might reasonably think that you already have an “AI” PC, because you can access services such as ChatGPT, Adobe Firefly or Google Gemini from your computer, making AI do creative work for you. While it’s true you’re using AI in those instances, the critical differentiation here is that all of the actual AI work is occurring in large data centres, with the end results being pushed back to your PC. Here the speed of AI generation is typically far more dependent on the speed of your internet connection (and to a lesser extent how busy those data centres are) than the speed of your PC. The data centres are doing the AI work, your PC is just receiving it based on your prompts.
So what does that have to do with “AI” PCs? For the new range of AI PCs, based around processors from Intel with its Core Ultra chips, AMD’s Ryzen 8000 chips or Qualcomm’s Snapdragon X Elite, the AI part comes from the integration of what’s called a Neural Processing Unit, or NPU. Neural Processing Units are dedicated hardware that can assist with generative AI tasks, as well as certain other factors such as video processing, which is why they’re being hyped for features like providing more realistic blur effects in video calls on Windows 11 PCs.
NPUs aren’t new in the hardware space, however – they’ve been present in the “Apple Silicon” processors running the past couple of generations of Apple Macs, and even before that, they’ve been present in a lot of smartphone hardware, being used primarily for photo processing effects more rapidly on-device.
In theory, having an NPU onboard could change the AI generation dynamic, shifting some if not all of the production work to your own computer. That should be faster, but also significantly more secure, because you’re not putting a task to an external server at all.
Copilot+ PCs promise a new consumer AI landscape
While AI has been a buzzword across all of IT in 2024, it’s only quite recently that the first consumer-grade laptops have started to come to market with the promise of onboard AI. While they’re technically Windows 11 PCs, they’re also being marketed heavily as “Copilot+” PCs.
Copilot is the name that Microsoft’s given to its onboard AI assistant, and if you’ve got a Windows 11 PC that’s had the latest updates pushed to it, you already have Copilot on board. It’s just that that version of Copilot all but entirely relies on making its AI queries to external servers.
Copilot+ PCs instead do much of this processing on the NPU within the processor itself. The first range of Copilot+ PCs will hit store shelves in late June 2024, all running on Qualcomm’s Snapdragon X Elite or Snapdragon X Plus processors. Those are ARM-based processors, distinct from the x86 architecture that’s been the heart of most Windows PCs historically. If you’ve got a PC running with an Intel or AMD processor under the hood, that’s x86.
ARM Windows isn’t new, but prior Windows ARM implementations struggled to run all Windows apps; Microsoft’s claim here is that the most popular apps are already ARM-ready, and for those that aren’t, a new emulation layer should allow older x86 apps to run as though they were proper ARM apps. We’ll have to wait and see how well that plays out, though it may become moot when and if Intel and AMD also bring out Copilot+ approved processors. That could happen quite rapidly according to some industry forecasts.
The big headline feature for Copilot+ AI PCs is called “Recall”; a backup and constant system snapshotting feature that’s designed to make it easier for Copilot+ users to find just about anything they’ve done on their systems, whether it was a snipped of a webpage, a song lyric or a document. It does so by constantly monitoring screen activity – which brings with it some not inconsiderable questions around privacy and security – but there is some scope there for improving productivity.
Copilot+ PCs will also leverage AI for live captioning and translations from more than 40 languages into English. Again, that kind of translation isn’t new – Google Translate does similar – but the difference here should be in speed, because it won’t have to head out to a remote Internet server to process audio, instead doing so on the laptop itself.
