The hype around artificial intelligence PCs is huge. In the days running up to the Consumer Electronics Show in Las Vegas held earlier this month, Microsoft declared 2024 “the year of the AI PC.” Before we jump on the AI PC bandwagon, it’s important to understand exactly what an AI PC is and what it can do for your computing experience — today, not in years to come.

The concept of an AI PC may sound revolutionary, triggering thoughts of a self-functioning computer while you sit back and relax. A few prompts, the right programs and a day’s worth of work could be produced in the background while you attend to other more interesting activities on your PC, checking in occasionally to make sure your AI assistant is on track. But we’re far from that utopia.

To qualify as an AI PC, a computer must have a new type of chip with a neural processing unit (NPU) such as Intel’s newest Meteor Lake chips and recent AMD chips. These new chips are designed to accelerate AI functions on the PC while using little power. But here’s the thing: All the AI-powered tasks we use today like ChatGPT and image creators run in the cloud, not locally on your computer.

There are a few small things that PC manufacturers have routed to the NPU like automatically reducing background noise when using the built-in microphone and adjusting the cooling fans inside your PC, but that’s about it. And Microsoft Windows 11 computers due to be released this spring will feature a new key on the keyboard to launch CoPilot, its AI assistant, but all it can do at this time is blur backgrounds on video calls and make your eyes appear to be looking at the camera when they’re not.

It’s fine to have a more sophisticated chip, but it’s a waste to pay more for a PC equipped with one until there is enough for it to process to make a difference. Experts estimate this may take several years.

When that time comes, the entire premise for consumer demand is that people will prefer to have locally based AI programs running on their PCs rather than accessing the cloud, or at least prefer a hybrid model — some cloud, some local. Manufacturers are betting on that sentiment, particularly from business customers.

“The ability to run generative AI applications locally will enable more personalized experiences, improve latency, reduce costs, and provide better security and privacy protections,” HP’s Enrique Lores said at Qualcomm’s Snapdragon Summit 2023 event in October.

Let’s unpack that statement. It’s true that over time, an AI assistant could learn and anticipate your computing habits, but if you deviate from your routine you may find yourself doing some time-consuming backtracking. Latency, the time it takes for your PC to respond to a mouse tap or keystroke, is important to gamers but not to regular users. How fast your computer replies to a ChatGPT prompt depends on several factors, and even today, responses come very quickly. Reduced cost refers to the business expense of using cloud servers compared with buying AI PCs for employees, a cost that will vary and does not apply to the home user, which leaves the last reason: better security and privacy.

But is it? Data stored in the cloud is usually encrypted, and anyone wanting to access that data needs to have the digital key. Not so for most personal computers. Big cloud computing companies store the same data in multiple places in case hardware breaks or part of a system is compromised. Can you say the same for your PC? To date, most of the biggest data breaches have exploited traditional computers at companies and personal machines, not data stored in the cloud.

Further, we’ve become accustomed to working in the cloud and understand the benefits. We know that if our PC fails, our documents will still be safe. It’s easy to share documents and collaborate on projects from a cloud-based service. We don’t have to buy a unit with the most powerful (and pricey) processor or the most storage. For the individual, convenience, peace of mind and cost savings are all benefits of working in the cloud rather than within the confines of our machines.

There will come a time when an AI PC makes some sense, especially for those who have embraced an integrated AI work system. This takes us back to the hybrid model mentioned previously where an AI chip can offer faster processing power while saving battery life, but the resource-intensive computing will still be done in the cloud. In about three years, there should be enough for the NPU to do, many AI PC models available from all major manufacturers and we’ll be several generations into AI chips — that’s the time to go shopping.

Leslie Meredith has been writing about technology for more than a decade. As a mom of four, value, usefulness and online safety take priority. Have a question? Email Leslie at asklesliemeredith@gmail.com.

Newsletter