All mobile phones are equipped with AI, but why do you disdain it?

Watch any mobile phone launch event and you will find AI features added. However, many users are somewhat dismissive of this and secretly think that it is just a hype. So, are these advertised features really AI?

[[227099]]

New selling point of mobile phones: AI

Since the second half of last year, almost all mobile phone products have been launched with the concept of AI. Samsung first launched the AI ​​assistant Bixby last year, and then emphasized that Bixby can complete intelligent translation, currency conversion and other functions through the camera when launching the Galaxy S9 this year. Huawei's Kirin 970 chip also integrated the NPU neural network unit for the first time, and the subsequent Mate 10 and P20 products have added the function of photographing objects. Among them, the AI ​​photography master of the P20 series can intelligently identify 19 categories such as animals, food, nature, portraits, etc., 500+ scenes, and then customize the photo mode for users through the calculation process of scene detection + recommendation rules + user habits = photo mode. The recently released Nubia V18 has also optimized the NeoSmart AI engine, indicating that the system can be more power-saving, smooth and easy to use.

Even Apple, which is relatively low-key, has added the word "bionic" to the name of the A11 processor, and also called it a neural network engine that can perform up to 600 billion operations per second. The use of the word "neural network" undoubtedly suggests to consumers that the chip also supports AI capabilities.

In a broad sense, it must be AI

"Damn, making smart photo album classification can also be considered AI." This is the comment of a netizen when he saw a press conference of a domestic mobile phone manufacturer. Indeed, now all mobile phone manufacturers have begun to move closer to the concept of AI, from chips to the above-mentioned photo album classification management. In fact, these functions are also familiar on previous mobile phones or computers, such as voice assistants, photo translation, location guidance using LBS+AR+camera, and mobile phone energy consumption control, which makes many people think that these functions that have existed in the past have been instantly packaged as AI.

So, what is AI? In addition to the literal translation of the four words "artificial intelligence", there is actually controversy over the definition of AI. For example, some concepts emphasize that AI needs to have the ability of machine learning and deep learning. Obviously, the intelligent function of some mobile phones that only uses the existing database and then makes corresponding judgments does not meet this concept.

In addition, in the concept of deep learning, some people believe that deep learning should be emphasized, which means that neural networks are needed to allow machines to simulate the human brain to learn, form logic by themselves, and then gain the ability to surpass the human brain. Just like Alpha Go completed a leap over human chess players after learning from chess records and playing against itself. Obviously, the AI ​​functions in many mobile phones are also difficult to do this. However, in this view, one problem is that humans currently have a very limited understanding of the human brain. 99% of the deep learning of computers is done in the computer's own way, which is somewhat far from the word "artificial" in AI.

However, AI has a broad concept that is currently more recognized, which is: taking reasonable actions based on the understanding of the environment and optimizing a defined objective function.

Obviously, all AI functions currently launched by mobile phone manufacturers meet this broad concept of AI. These functions are based on the user's environment and then automatically take corresponding actions. It can also be said that even Siri, some battery management applications, and instant translation applications that users are already very familiar with can be called AI.

There is a gap between AI level and user expectations

However, why do the above users still complain about the abuse of the so-called AI concept? The obvious reason lies in the current industry's detailed discussion on the concept of AI. The Alpha Go case seems to have given the world a new revelation, that is, the originally weak AI can reach a level that exceeds the human brain, which also gives AI a new label.

The super learning ability behind Alpha Go has also become a new symbol of AI. Machines can not only rely on existing big data to respond, but can also learn by themselves through databases, find the logic by themselves, and deduce new solutions. This is a new development direction for AI.

However, for AI functions that do not have autonomous learning capabilities, since they are not much different from the existing functions that users have been exposed to, it is difficult for users to make a strong association between them and the currently popular AI concepts (emphasizing neural networks and deep learning). Naturally, there are differences with their inner expectations, and complaints are inevitable.

Of course, if AI's self-learning ability is applied in mobile phones, it also involves the privacy issue that is increasingly valued by users. If the data collected by AI during the learning process is stored in the cloud, it will obviously cause huge controversy. For example, after long-term learning, voice assistants and face recognition can obviously reproduce the user's appearance and voice, which will cause great risks. Therefore, Apple and Huawei have also built-in special neural network units in their mobile phones to store such data, but they will also face the problem that if the user changes the phone, the machine needs to re-learn the user's characteristics. Only after the user has used it for a period of time can the machine reach the original learning level.