In 2023, it is not the iPhone 15 or the Huawei Mate 60 that will be popular in the tech circle, but ChatGPT. The explosive growth of AI large language models has given rise to many practical functions, such as AI drawing, AI speech, AI copywriting, etc.
With the explosion of AI large language models, mobile phone manufacturers have also begun to access self-developed AI large language models in their mobile phone voice assistants. Like OPPO's AndesGPT, vivo's blue heart model, Xiaomi's milm, and Huawei's Pangu model. On the side of international manufacturers, Samsung barely kept up with the rhythm of domestic manufacturers' involution, and disclosed its self-developed large model - Gauss in the middle of last year. However, Apple, which won the 2023 global mobile phone market sales championship, was indifferent, which will inevitably be questioned by the outside world.
Apple's AI, always one step behind?
After all, the old rival Samsung has added more AI functions to the GalaxyS24 series, and the iPhone 15 series and iOS 17 have not shown AI-related functions to the outside world, is Apple really lagging behind in the field of mobile phone AI? Not really. In the Steve Jobs era, Apple was arguably the first technology company to deploy AI on mobile phones. As the last generation of products that Steve Jobs participated in, the iPhone 4S introduced the concept of intelligent voice assistant for the first time, but at that time, Siri was limited by technical reasons, and Siri did not provide much convenience to users, and even normal voice communication was difficult. However, Siri did set off a wave of craze at the time, and many Android manufacturers also launched their own mobile phone voice assistants.
Looking back at Siri back then, it seems that it is not an exaggeration to see it as the prototype of the current AI large language model. Although in the early days of mobile phone voice assistants, Apple fired the first shot and took the lead in launching voice assistants on mobile phones, but later due to problems such as vague positioning and chaotic management of Siri's internal team, Siri was eventually surpassed by Android manufacturers.
Siri, once a glamorous company, has now become a laughing stock in people's eyes. In the public perception, the label "artificial intellectual disability" has long been associated with Siri**. It is undeniable that Apple was the first manufacturer to add a smart voice assistant to its mobile phone.
In the Cook era, Apple's exploration of iPhone AI-related features has not stopped. Take the iPhone X, for example, which is Apple's first full-screen phone and the first phone to be equipped with Face ID. If you want your phone to accurately recognize face information, you need a deep combination of software and hardware. The iPhone uses the Depth Sensing Camera to project invisible light from the human eye to capture face information, and builds a depth map and infrared image of the face, and then uses the neural network engine of the A-series bionic chip to compare it with the previously registered face information to confirm the user's identity.
That said, starting with the iPhone X, Apple has been using AI to help machines identify and confirm the identity of users. At the same time, Apple is also the first manufacturer to use structured light technology to unlock mobile phones.
Therefore, Apple's technology in mobile phone AI is not lagging behind, and it is even one step ahead in some AI features. However, it is an indisputable fact that the current iPhone is not equipped with AI large language models.
Generally speaking, whether a certain technology is backward or not is often not based on the chronological order of technology application, but on the quality of the experience after the technology is implemented. For example, one brand of mobile phone is equipped with AI large language models earlier than another brand of mobile phone, and they all support AI drawing. When you ask them to draw a short English short, you find that the phone that released the AI large language model later successfully drew the English short, and the mobile phone that released the AI large language model earlier drew three children.
In terms of experience alone, the mobile phone that releases the AI large language model later is better than the mobile phone that releases the AI large language model earlier. Therefore, it cannot be said that the iPhone is not equipped with an AI large language model, which means that Apple's technology in this area is lagging behind. **Whether Apple's AI large language model technology is lagging behind will need to wait until the iPhone implements this feature before making a judgment.
Applegpt is ready to go
For tech companies, technology is a must-have. And technology giants like Apple, AI large language models and other technologies are naturally well prepared. Friends who know Apple should know that Apple has high requirements for the application of technology, either waiting for the technology to mature, or polishing the technology well enough, otherwise it would rather be abandoned than affect the user experience.
Previously, Apple whistleblower Mark Gurman said that Apple's head of software engineering, Craig Federighi, told his subordinates last year that the operating system, codenamed Crystal, would usher in a revolutionary upgrade. Mark also said that Apple internally defines iOS 18 as an epic upgrade.
In an interview with Bloomberg, Apple CEO Tim Cook said that Apple has been working on artificial intelligence technologies, including generative AI, and has invested heavily in R&D for it over the years.
It is not difficult to see that Apple does have the technical reserves for AI large language models. Various leaks also indicate that Apple will announce Apple's self-developed AI large language model at this year's WWDC and will apply it to iOS 18.
Combined with the current situation, the author believes that Apple has a high probability of announcing its own AI large language model at WWDC 24 and will integrate it into Siri. Because almost all of the Android camp has launched its own AI large language model, even Google has launched its own AI large language model Bard.
Even Microsoft, Apple's best developer, has launched its own AI large language model Copliot by connecting to GhatGPT. Not only that, but technology Internet giants such as Meta and iFLYTEK have also launched their own AI large language models.
Major technology and Internet giants are taking advantage of the popularity of ChatGPT to build momentum for their own AI large language models, and at the same time further seize the market. As the world's largest technology giant by market capitalization, Apple will naturally not give up on the AI large language model market.
Self-developed AI large language model, Apple has long been premeditated?
In fact, we can see from the A17 Pro chip of the iPhone 15 Pro series that Apple may be preparing for the upcoming AI big model. Regarding the A17 Pro chip, almost all of them are negative comments on the Internet, such as the A17 Pro, which insults the name of Pro, and is rubbed on the ground by MediaTek Dimensity 9300 and Qualcomm Snapdragon 8Gen3; Or Apple's A-series chips fell off the altar, and the performance was surpassed by friends, and TSMC's 3nm process technology was just a gimmick.
Truly. The theoretical performance of Apple's generation of chips is not very good, and the CPU and GPU are only slightly improved compared with the previous generation. However, the number of transistors in the A17 Pro stacks up to 19 billion. And Apple's desktop entry-level chip M2 has an internal transistor count of 20 billion.
Comparing the two, it can be found that the number of transistors in the A17 Pro, a mobile phone chip, has reached the scale of a desktop-level chip. This begs the question, why does Apple stack so many transistors on the A17 Pro? The author believes that Apple is likely to pave the way for the future AI large language model, so as to realize the offline local operation of the iPhone.
Since the iPhone 13 series, Apple has distinguished between the chips equipped with the standard version and the Pro version of the iPhone, and the volume model will be equipped with a castrated version of the chip, while the flagship positioning of the Pro version will be equipped with a full-blooded version of the chip. When it comes to the iPhone 14 series, Apple has changed its strategy again, and the Pro version of the iPhone will use the latest chip, while the standard version will use the chip of the previous generation Pro model.
If you extrapolate according to this logic, this year's iPhone 16 series standard version will continue to use the previous generation 17 Pro chip, while the iPhone 16 Pro series will use the A18 Pro chip. If Apple adds its own AI large language model to iOS 18 and integrates it with Siri, then even the iPhone 16 standard version can realize the offline local operation of AI large language models, after all, the transistor scale of the A17 Pro chip has reached the desktop level.
The iPhone 16 Pro series is equipped with the A18 Pro chip, and its internal transistor number and performance will be further improved compared with the previous generation. Therefore, compared with the iPhone 16 standard version, the Pro version may be better at running Apple's AI large language model offline.
Generally speaking, the later a vendor releases an AI large language model, the more advantageous it is. Because if you want AI large language models to be more intelligent, it takes a lot of time to train. Take ChatGPT 40, for example, has reached 1 trillion parameters at present. And vivo's blue heart small V, its parameter volume is 7 billion.
Apple is the latest entrant to the game, not only can it learn from its opponents, but it can also spend more time training AI large language models with a larger number of parameters. At the same time, it can also introduce new core selling points for the next generation of iPhone, further stimulating consumers' demand for replacement. For Apple, it is undoubtedly the best way to kill three birds with one stone.
In the future, in addition to volume images, AI large language models may become new involution objects for mobile phone manufacturers. The more involution mobile phone manufacturers have, the more profit we consumers will make, and the better the experience of using AI large language models on mobile phones, which can bring more convenience to users.