Apple moves towards the AI new era: chips, applications, and large models fight on three fronts

At the 2023 Developer Conference, Apple CEO Tim Cook unveiled the new mixed reality headset Vision Pro. However, Apple remained silent on the then-popular Generative AI.

A year later, Google and Samsung introduced the new concept of "AI Smartphones" with the Pixel 8 and S24 phones. Companies like Microsoft, Meta, and Amazon invested billions of dollars in hardware for Generative AI, while Apple seemed to still be "keeping silent," raising concerns about whether Apple would miss the opportunity for technological transformation.

However, with the highly anticipated WWDC 2024 approaching, Cook, once again taking the stage at the developer conference, has the opportunity to change this situation. The market expects Apple to release the iOS 18 operating system and a series of AI services and features, marking the beginning of its Generative AI plan.

The media believes that Cook is under tremendous pressure as Apple transitions into the AI era, facing the daunting "three-pronged battle"—

On one hand, they need to develop chips that increasingly support AI functions on their devices, while on the other hand, they need to create so-called "killer" applications to attract consumers. At the same time, they need to ensure access to the most advanced Generative AI models controlled by competitors Microsoft and Google.

Morgan Stanley analyst Samik Chatterjee stated:

"Apple needs to dispel the perception in the market that it is lagging behind in Generative AI and prove that it has caught up with other companies."

1. Powerful Chips, Memory Challenges

Regarding chips, the media believes that Apple already has powerful enough chips to run "AI Smartphones." Since 2017, Apple has had dedicated neural network architectures, with its latest A17 Pro chip exceeding 30 TOPS on the iPhone 15 Pro and Pro Max.

These NPUs with neural network engines make it easier to run extremely computationally intensive AI applications locally on devices. However, they also bring "significant technical challenges in terms of memory," as noted by Reece Hayden, an analyst at technology intelligence company ABI Research:

"Even very small Generative AI models have higher memory requirements than any current phone."

Currently, the iPhone 15 Pro only has 8GB of memory, while the Samsung S24 smartphone powered by Qualcomm chips has 12GB.

Research firm Counterpoint estimates that "AI Smartphones" will account for 43% of global smartphone shipments in 2027, with a usage of up to 1 billion units.

2. Battle of Applications and Large Models

In terms of applications, the media believes that one practical measure Apple can take is to enhance Siri using AI.

Tim Bates, a professor at the Innovation and Technology Institute of the University of Michigan-Flint, stated that within the constraints of its existing hardware, Apple can leverage the personal data stored on devices to create personalized experiences for its users.

This so-called "on-device" approach also has an additional benefit of protecting user security and privacy, as consumers are less likely to want AI applications to train on their personal information and export it to the cloud. Running functions locally can also eliminate the latency involved in generating responses from remote servers. Bates added:

"Siri is really the perfect 'flavor' of interactive AI... individuals can control their data, converse with AI, complete tasks, without worrying about data being siphoned off."

Vamsi Mohan of Bank of America believes that the killer application of AI on the iPhone will emerge when Apple can provide a "conversation assistant that is fully integrated with everything on the device, and then interacts with some kind of expert agent in the cloud."

This is also where the market expects Apple's partnership with companies like OpenAI to come into play. While Apple has been developing its own generative models, it cannot match the state-of-the-art models based on tens of trillions of data parameters, hence the need to collaborate with other companies.

These initiatives may put Apple on par with some competitors, but have not yet surpassed them. The specific contours of any agreement between Apple and OpenAI in terms of partnership and progress remain unclear. Media analysis suggests that whether these two companies have cultural alignment and how long this "marriage" can last are still pending questions.

However, experts caution that it is too early to count Apple out. The AI revolution is still in its early stages, and no major tech company has successfully applied generative AI to killer hardware products. Jason Banta, Corporate Vice President of AMD, stated:

"To use a sports analogy, I think we are still in the first minute of the first quarter... It's still too early now."