Lost amid today’s hoopla over a pair of brand new M3 MacBook Air laptops was a none-too-subtle shift in product messaging that might mark Apple’s official entry into the race to build an AI PC.
In the press release about the new 13-inch and 15-inch MacBook Air ultraportables with the latest Apple silicon, two paragraphs were devoted to the claim that the MacBook Air is the “World’s Best Consumer Laptop for AI.” If you didn’t follow the computer industry like I do, you might have attributed that to some strangely specific boasting or hyperbole on Apple’s part. Although I see it a little differently.
First a little history. Until 2020, virtually all new Apple Macs, including MacBook Airs, had Intel CPUs. That year, Apple announced its intention to build its own chips and eventually replace all Intel CPUs, with a custom system on a chip (SoC) that became known as Apple Silicon. The first such chip – the M1 – arrived on the beloved MacBook Air M1 (now discontinued).
The rise of silicon
Apple eventually made good on its promise and through multiple iterations and upgrades, eventually replaced all of the Intel silicon with its own. Intel still owns most of the world’s worth of Windows PCs, but in some ways Apple is seen as the system innovator, creating SoCs that run faster and more efficiently than anything Intel can produce.
Intel’s big plan to counter that perception, and excite people looking for other ARM-based solutions that can run Windows as quickly and efficiently as something like Apple Silicon could, is to overhaul its entire chip line. revised with Intel Core Ultra processors and, most importantly, the ‘AI PC’.
The AI part comes from the Neural Processing Unit or NPU which will work with Intel Core Ultras as a kind of AI coprocessor. Intel has the support of virtually all major Windows PC manufacturers and, perhaps more importantly, Microsoft. The Redmond-based software giant is in the midst of a full Copilot press. The generative AI, formerly Bing AI chat, which is built with the intelligence of OpenAI, is seemingly everywhere, and on AI PCs it will appear as a Copilot keyboard button.
What exactly any of us will do with an “AI PC” is unclear, but we’ll be talking about these systems all summer long and into the Northern Hemisphere’s back-to-school shopping season.
By some measures, Apple owns it only 17% of the PC market. Even if people believe that Apple silicon is better and macOS is a better platform than Windows, they can’t afford to sit back and let Intel and Microsoft innovate and market themselves to even greater PC market heights.
We know AI
This brings us back to the ‘world’s best laptop for consumer AI’. However, Apple has a point. It has been doing AI for ages, going back to adding the first Neural Engine to the iPhone 8 courtesy of the A11 Bionic CPU. You can draw a direct line from this early built-in machine learning system to the 16-core neural engine on the M3.
Apple has made no secret of its silicon’s built-in AI capabilities, but it has never made it front and center. That all changes now.
The company has no choice. Part of Apple’s problem is that, unlike Microsoft, OpenAI (ChatGPT), and Google (Gemini), Apple doesn’t have a significant generative AI product. Siri is not generative, it cannot generate poetry, presentations or works of art. That has hampered Apple’s efforts to lead the way.
However, in the release, Apple specifically mentions Large Language Models (LLMs): “Combined with Apple Silicon’s unified memory architecture, MacBook Air can also run optimized AI models, including Large Language Models (LLMs) and diffusion models for local image generation with Well performance.”
The shape of things to come
Running locally, without the need for potentially less secure or slower cloud support, has always been Apple’s secret AI sauce. However, Apple is fully aware that it cannot win this game if it does not allow cloud-based generative AI.
In demonstrations I saw, I saw that the MacBook Air M3 could perform both cloud-based Microsoft CopIlot prompts and local generative tasks with apps like Luminar Neo, which can take a blurry nighttime photo and add generative information to make the image usable. In both cases, their performance seemed almost instantaneous and easily as good as what you’d get from a cloud-based generative AI.
However, the purpose of showing off these apps and making these statements isn’t just to tell the world that Apple is also doing Gen AI. I think it prepares us for what’s to come.
It’s not just new products and press releases. Apple CEO Tim Cook now takes almost every opportunity to promise big things in generative AI (remember when he talked about ‘AR’? What a difference a letter makes).
Cook knows that Apple silicon is more than ready for big language models and generative AI for images and text, and we’ll see Apple take advantage of all that power starting with WWDC 2024 in June.
That’s Apple’s message when it comes to AI: you ain’t seen nothing yet.