Apple doesn't hype artificial intelligence? they just didn't make a mockup

Source/ Big Model House

Author/ Wang Haoda

Image source: Generated by Unbounded AI tool

At the just-concluded WWDC 2023 conference, Apple once again chose to "treat" the current artificial intelligence with caution. Compared with "Artificial Intelligence" (artificial intelligence), Apple seems to be more inclined to use "Machine learning" (machine learning), To restore the essence of technology...

Apple first mentioned the word "Machine learning" in the WWDC 2017 keynote speech on June 5, 2017. At that time, Craig Federighi, Apple's senior vice president, announced two machine learning frameworks, Core ML and Create ML, to make it easier for developers to integrate machine learning models on Apple devices. Also after this WWDC, Apple has been continuously promoting the development and application of machine learning technology.

For most consumers, when it comes to Apple's artificial intelligence, the most impressive application is Siri, which is "full of bugs". Since the debut of the iPhone 4S in 2011, it has become the object of everyone's "teasing". Until now, Siri's communication ability is still not at the forefront of the market, and it has become a tasteless function that few people care about. Faced with the blowout growth of the artificial intelligence industry, Apple, which has been slow to make new moves in the field of artificial intelligence, has made the industry generally believe that it has begun to enter the ranks of "left behind".

Firmly based on a corner of the market, Apple prefers "precise"

In the face of various voices of doubt, Apple CEO Cook said that Apple is still a hardware manufacturer in essence, and the company does not have the pressure to improve productivity by upgrading software like Google or Microsoft.

In WWDC 2023, Apple also launched the iOS 17 system, which provides users with a series of new features and improvements, including personalized and interactive features for calls and FaceTime, a new diary app, a new standby mode, and new dynamic widgets. Tools, enhancements to the CarKey digital car key functionality, improvements to the ARKit API framework, and more.

Relying on simple, high-quality, personalized products and services, Apple has won a large number of praise and loyalty. In the mobile market, with its market share and the advantages of a closed and efficient iOS ecosystem, it has attracted more users. At the same time, Apple continues to innovate and expand its product line, creating a series of self-ecological products such as Mac, iPad, Apple Watch, AirPods, and Apple TV.

In short, Apple's strong brand appeal and ecological influence make its consumer electronics products (except for luxury brands) the same kind of "ceiling" in terms of pricing, but still firmly occupy a corner of the market in terms of sales. Strong product advantages are Apple's unique confidence, but the strong appeal of developers is also one of the factors that its App Store continues to contribute revenue to Apple, which means that developers' demand for artificial intelligence continues Today's growth, Apple's cautious attitude does not mean that it does not face up to the future development and prospects of this technology.

"Curve to save the country" service hardware, Apple just didn't make a big model

As a hardware manufacturer, Apple has its own chips, services, systems and other products. The complete ecology formed by these products also provides Apple with a strong infrastructure and platform in the field of artificial intelligence, which can achieve efficient collaboration between hardware and software. And optimization, improve user experience and performance, and better protect user privacy.

The upcoming update of iOS 17 also uses the Transformer language model, which can predict input words based on user semantics and language habits, and make automatic error correction more accurate; in terms of speech recognition, iOS 17 uses a neural engine to make speech recognition more accurate.

It is worth noting that the risk of data and privacy leakage in the field of artificial intelligence and the supervision of artificial intelligence have become issues that the world must face together. A series of policies and regulations have been introduced at home and abroad to maintain the development of artificial intelligence. Even Sam Altman, the father of ChatGPT, also applied to the government for supervision of OpenAI at the hearing.

Mr. Bao Linchao, head of Unity China's AI technology, told Big Model House: Artificial intelligence is hungry for data. At present, issues such as data and privacy security have delayed the development of artificial intelligence to a certain extent, which requires enterprises to pay close attention. The problem of data access is to seek ways to promote the progress of artificial intelligence on the basis of ensuring that user privacy is not violated.

And Apple's emphasis on user privacy and data security is well known in the industry, and even regards its privacy protection function as an important selling point of the product. Apple has adopted a variety of technical means, such as encryption, biometrics, differential privacy, etc., to protect user data from being leaked or misused. At the same time, this will enable Apple's artificial intelligence business to access and use user data more compliantly, serve users more accurately, and greatly improve the response speed and reliability of artificial intelligence.

Vision Pro performs the art of stacking

In the upcoming Vision Pro product display, it is not difficult to find that although Apple did not mention "artificial intelligence", it seems that all new actions revolve around "artificial intelligence".

Vision Pro also uses a Transformer-based speech recognition model to make speech recognition more accurate. And it can effectively process sequence data of natural language. As a deep learning model based on the self-attention mechanism, it can also perform automatic error correction and word prediction, improving the accuracy of speech recognition.

At the same time, Vision Pro also relies on visual generative modeling to scan people's facial information through the front camera, and then based on machine learning technology, the system will use advanced coding neural networks to generate a "digital clone" for the user. Using advanced machine learning models, based on the user's eye tracking and pupil status, to predict the user's body and brain status, such as whether you are curious about the current thing, whether you are distracted, whether your attention is distracted, etc.

This data can help Vision Pro provide a more personalized and intelligent experience, such as updating the virtual environment based on the user's attention, relaxation or learning, or creating biofeedback based on the direction of the user's eye gaze.

There is also a convolutional neural network equipped with Vision Pro, which achieves efficient and accurate image recognition and object detection functions through convolutional layers, activation layers, pooling layers, and fully connected layers, providing users with rich and realistic mixed reality objects. , It plays an important role in the fields of automatic driving, security monitoring, medical image analysis and so on.

It is worth mentioning that Vision Pro is also the only MR headset on the market that can be controlled without a handle at all. Therefore, if you want to realize human-computer interaction through dynamic capture, dynamic analysis, eye tracking and other technologies, it puts forward extremely high requirements on the computing and processing capabilities of Vision Pro.

Vision Pro has two built-in chips, M2 and R1. The M2 chip is responsible for running the operating system and application programs, while the R1 chip is responsible for functions such as positioning, collaboration, visual image processing or transmission.

The R1 chip is Apple's new chip designed specifically for Vision Pro. It is dedicated to processing sensor data and spatial calculations. It can stream new images to the display in 12 milliseconds, 8 times faster than the blink of an eye. An immersive mixed reality experience is critical. It can be said that in order to realize machine learning to meet the needs of users, the R1 chip is the "answer sheet" that Apple handed over to the world in the era of artificial intelligence.

For Apple, perhaps the word "Machine learning" more accurately describes Apple's technical characteristics. But through Vision Pro, iOS 17, and looking at the entire Apple ecosystem, it is not difficult to see that Apple does not want to vaguely define artificial intelligence. They are keen to use technology to transform or expand their product categories and build their own brands, rather than using industry-wide vocabulary.

It is true that, compared to large artificial intelligence models such as ChatGPT, Apple's accumulation in the field of artificial intelligence does not allow it to have the ability to "change the world" like the iPhone, but Apple uses the Vision Pro product to try to educate the world How to combine AI with the scene of spatial computing to achieve the effect of "1+1>2". Using another form of XR implementation (the previous one was Microsoft's HoloLens), opened the door for the industry to enter the "spatial computing era".

Compared with companies such as Microsoft and Google, Apple is less open to artificial intelligence technology, but relying on the support of huge market demand, Apple hopes to maintain the high-quality experience that its products bring to users, whether it is "machine learning" or "artificial intelligence". "Its core essence is still what it can bring to users. Big Model House believes that whether it is hardware or software users only care about whether the product can perfectly match their needs when using it, how high the computing power and how many data parameters ultimately score the product can only be the satisfaction of consumers To a certain extent, I believe that the final form of artificial intelligence development will move in the direction of civilianization and dailyization.