The reason for this is simple. Mobile phone product development cycle is usually 1-2 years, product development phase does not join the AI-related plans, but the AI concept is too hot when the competition is nearing, and competitors continue to emphasize the AI concept. To force one to have to do so at the time of the promotion, this has led to the so-called AI mobile phones being caught in an all-or-nothing situation.
The AI proposition is undoubtedly attractive to the entire mobile industry. Like the 5G, it will become an important constant for the mobile industry to change. However, there is a set of methods and logic for identifying genuine and fake AI mobile phones.
Three routines of fake AI phones
The 2016 AI concept has begun to take off, and there have been mobile phone manufacturers 蹭AI hot spots. Throughout the past two years, almost all the routines of fake AI mobile phones are consistent.
1, intelligent process management called 'AI intelligent scheduling'
After the Android 6.0 upgrade, the Doze mode was added. This is the 'living mechanism' of the IOS 'tombstone mechanism'. The so-called 'buried mechanism' means that after a user has not used a mobile phone for a while, he will suspend network access and delay. Operation, Synchronization. Over time, as the device is not recharging and is stationary for a long time, the system schedules fewer and fewer sessions, helping to reduce battery consumption.
We can see the logic of the Doze pattern from the diagram of the development of the siege lion by the Android framework called “Fireflies”. This is a mechanism for the Android system to perform resource recycling calls based on runtime.
After Google launched Android 6.0, it warned all mobile phone manufacturers not to emasculate this model. That is to say, all mobile phones equipped with Android 6.0 actually have this smart scheduling mechanism.
This is a natural feature of Android 6.0. However, it is used by some mobile phone manufacturers as their selling point. Changing the name of 'AI intelligent scheduling' is used to sell. This routine often misleads the outside world to think that mobile phone manufacturers have developed their own intelligence using AI. The function of scheduling.
But the real smart scheduling actually needs to cooperate with the AI processor, learn user behavior, and decide how to schedule the memory according to the daily use of the user. These resources of the processor are more efficient in the use of resources than the Doze mode. This problem is for the time being. Press not a table, the following will be detailed in the real AI phone.
2, a single software AI function exaggerated into 'AI mobile phone'
AI itself is a very rampant concept. The voice assistant in mobile phones is AI, the depth of field on the camera, the aperture algorithm is AI, the face recognition is unlocked, and even the intelligent skin of Mito Xiuxiu can be AI, today's headline gives you Watching the news mechanism is also an AI algorithm - AI is not unusual, and it is everywhere in our daily life.
We already had a lot of good AI landing apps on our mobile phones. It wasn't called AI before. It was called an algorithm-like portraiture of a smartphone. This is a typical application that depends on algorithmic imaging.
These software AI functions can indeed be regarded as AI landing on mobile phones, but it should be noted that these technologies are only some of the more basic applications of artificial intelligence technologies. Mobile phones equipped with these features are absolutely not AI mobile phones.
There are a large number of mobile phone manufacturers. After releasing these functions at the conference, they dare to boast of AI mobile phones. This practice is purely drowning marketing. However, for some beautiful white people, it looks like it is in the fog. New technology, this kind of routine has become a popular means of fraudulent users through information asymmetry in the Xiaobai market.
3, The AI function is added to the chip and called 'AI chip'
Some chip manufacturers, such as Qualcomm, will add a machine learning framework to the chip to improve the performance of the camera's computing. This feature was added to all high-, medium-, and low-end chips in 2017. In other words, Qualcomm 835, Qualcomm 660 and Qualcomm 625 These high school low-end chips all contain this technology.
In fact, this technology does not have much novelty. Originally, this was a very regular action, but it was often interpreted by handset makers at the press conference as deliberately interpreting it as an 'AI chip', and then deducing that it was their own production. AI mobile phone.
This routine and Android 6.0's Doze model is almost exactly the same, both of which are deliberately exaggerating the false propaganda of the new technology of the upstream manufacturers. Such marketing routines are even used as words in offline stores. It's true that bullies don't understand AI.
In general, a fake AI mobile phone is just an ordinary mobile phone. For marketing, it is affixed with AI elements. For consumers, it does not have much improvement, and it is disrupted in the market. Voice. Sometimes consumers even have to pay more for the so-called 'AI'.
Three Dimensions of Real AI Mobile Phones
For ordinary users, the mobile phone is a mobile phone, and it is really practical. The safe AI is a good AI. How do you see through the various AI 忽悠 忽悠? In fact, you can recognize the true AI flagship through three dimensions.
1, to have an independent AI processing unit
Last year, some media questioned the use of neural networks in artificial intelligence processors. In fact, this technology has become more popular. Adding a processor dedicated to complex data sorting algorithms can help smart phones and other technologies become better, and Various new technologies provide better support, from automatic image enhancements to faster video library searches.
Mobile phones without AI processors are not really AI phones. The most typical AI processor on the market is the A11 bionic smart processor on the Apple iPhone X. It has a chip with an independent AI processing unit. The neural network engine Joining allows the iPhoneX to distribute tasks properly between the performance core and the performance core.
For ordinary people, the most intuitive benefit is that the battery life is greatly improved. It will distribute tasks reasonably between the performance core and the energy efficiency core, intelligently managing the performance of the chip. Whether it is texting or browsing the web, the main chip will reduce the computation. Amount, increase battery life.
In China, the Unicorn 970 on the Glory V10 is also an AI chip. Utilizing the NPU specifically to handle machine learning-related operations, HiAI mobile computing architecture, AI performance is significantly better than CPU and GPU. Intelligent management and balance of performance at the core can be performed intelligently. , According to the use of state to adjust the performance of the processor core, so as to achieve fast operation, power saving purposes.
2, with supporting AI operating system
The AI operating system is also an integral part of true AI mobile phones. With AI chips, 8-16-bit floating-point numbers, pattern matching, database lookup, bit-field operations, and highly parallel processing, all can speed up completion. Instances.
Android 8.0 released in 2017 is actually trying system-level AI functionality. For example, in the face of Android's ills - Caton, power consumption and other issues, Android 8.0 is using practical AI to help users intelligently determine the frequency of use of software. Time-selective retention tasks. For some software that is not commonly used, the system will automatically shut down the software, which can greatly reduce the burden on the CPU and make the phone smoother and lighter. It is said that on the upcoming Android 9.0, Google will also be in voice And explore AI on two dimensions of the operating system.
In fact, the earliest domestic attempt to implement AI at the system level is Huawei's EMUI. The EMUI 8.0 on the Glory V10 is also a typical AI operating system. The system understands and summarizes user habits, under the guidance of the 'critical business priorities' rule, to users. Frequently used software and custom-enabled software at certain points in time open up higher computing privileges.
The EMUI8.0 system of the Glory V10, together with the AI processor, can achieve a reasonable allocation of the overall hardware resources of the mobile phone, perform multi-core heterogeneous resource scheduling, enable application resource-aware scheduling, greatly improve system fluency, and remain smooth for 500 days of long-term use. - This changes the Karen's dysphagia in one year for Android phones.
Even compared to mobile phones equipped with the Qualcomm 845 processor, the Glory V10 still does not fall out of the air in the case of grabbing red packets. This is actually the role of AI.
And native Android 6.0, 7.0 on the Doze mode of the kind of stiff screen-down time scheduling memory mechanism. This is based on user behavior analysis, and then intelligent scheduling resources approach.
3, have open bottom AI and AI ecology
Currently domestic smart phones wading AI, most need to force or construct AI ecology. Because of the mobile phone system, those AI functions carried on the hardware are not enough, they can only make the phone faster on a mobile phone. , More energy-efficient, more convenient, but it did not form a complete ecosystem in conjunction with more software vendors, causing the smart phone to move toward the next stage of chemical reactions.
Artificial intelligence technology needs the participation of developers, and developers can form a complete ecosystem. Open AI ecology also means that more APP developers can actively adapt to AI chips in the future, have higher efficiency, and develop more use. Scenario - In the future, with the landing of 5G networks, users can use AI voice to interact with applications in the system.
Huawei has already tried its own platform, such as opening the Unicorn 970 as a mobile computing platform, allowing more developers to connect applications to Huawei's AI platform. This makes mobile phones more affordable for shopping, identifying destinations, real-time travel encyclopedias, and travel. Assistant, travel assistant, automatic entry into driving mode, learning evolution, adaptation to the owner's habits and other functions.
This ecology will attract more application developers to come together and provide more diverse and rich AI products for the system.
Make the best guess - the future mobile phone will never be offline in the 5G network environment. There is no need to download the application. Each application will be able to provide a wide range of suitable activities according to the user's geographic information at any time because of AI scheduling. service.
We can even link to the few days ago Huawei, Xiaomi, ZTE, Lenovo's 'fast application'. Users do not have to download and install, click and play, can enjoy the performance experience of native applications - this may be with AI and 5G There will be a big relationship.
Because one APP creates an island of information, the logic of 'fast application' can facilitate data interaction between the chip, system and 'fast application', providing users with more accurate services.
Of course, this is an ideal state. It also requires mobile phone manufacturers, chip makers to share the same enemies, to make more 'fast application' of such a unified interface, the standard alliance, the future AI can land on all smart phones.