Table of contents
1. The application of AI in photography
2. Application of AI in daily life
3. Where does the AI on mobile phones come from?
Although 2023 is only halfway through, the technological shock from the AI field has no intention of ending. Looking back on the past, AI has never been as popular as it is today. It seems that it has become a new technological direction overnight.
Of course, the AI that everyone is talking about now is more represented by Chat GPT, a natural language processing tool driven by artificial intelligence technology. It can talk, draw pictures, and be competent for various jobs; it understands logic, is good at reasoning, and can even interpret people's emotions and attitudes.
1. The application of AI in photography
In fact, when it comes to AI, most users are familiar with it. We keep our mobile phones in our hands every day, and many of the first people to come into contact with AI applications come from smartphones, the most representative of which is the mobile phone image.
In the process of taking pictures on mobile phones, the role of AI can be called immediate. Taking Mi 13 as an example, when we enter the camera interface, there will be an option for "AI camera" in the drop-down menu item.
After turning on the AI function, we will find that the color of the photo is brighter, the saturation is higher, and the brightness also changes. With this function, when you press the shutter, the mobile phone will automatically calculate the optimal solution according to the shooting scene at this moment, so as to directly "P" a finished film, saving the post-production steps.
It is obviously not enough to just do this. With the changes in the AI capabilities of smartphones, it can be said that the scenes involving AI participation in mobile phone images are getting deeper and more frequent.
Among them, the night scene is a very typical scene. As we all know, the larger the area of CMOS, the more light will enter, and the better the image quality of night scene photos will be. However, limited by the size of the body, it is difficult to fit a larger sensor and optical lens inside the phone. The more complex the light, the image is prone to noise, which looks blurry, and it will become more difficult for the phone to shoot night scenes.
But in fact, we found that the mobile phone is actually very good at shooting night scenes, not even losing to some professional equipment. Why is this so? The answer is clear at a glance, and it is naturally related to AI.
With the participation of AI, the night scene photos taken by the mobile phone are no longer a single exposure, but multiple photos under different parameters are taken continuously, and then stacked and merged into a sample. This process is still relatively complicated, but using AI can quickly find out the feature points in the proofs for calibration, and then synthesize multiple frames to effectively remove noise and improve picture details.
In addition to basic applications such as AI mode and night scene mode, with the improvement of AI capabilities, advanced applications such as tracking everything, fast capture, magic sky change, and background blur have also been derived, which greatly enriches the gameplay of mobile phone cameras. Improved shooting experience.
All things focus
Just like tracking everything, the stable, precise, intelligent and sustainable focus tracking ability is especially suitable for shooting moving objects. If you have lively children or pets at home, this function is not very easy to use.
magic change
There is also Magic Sky Change, you can edit all kinds of sky effects for the photos in the album. It doesn’t matter if you can’t take a picture of star trails. You can generate it with one click in the later stage. , and even one-click to add lightning, sunset, aurora, galaxy... Not only is the gameplay novel, but it also seems to have no sense of disobedience.
2. Application of AI in daily life
So in addition to taking pictures, which scenes will use AI in daily use?
In fact, there are quite a few. For example, in Xiaomi mobile phones, there is a very interesting AI application, which can be said to be one of my favorites. It is AI calling.
When a call comes in, we can't answer it, but we don't want to miss it, or we are always annoyed by marketing calls. In this case, "Xiaoai Call" can help us complete the "answering" work well.
For MIUI users, everyone is no stranger to Xiao Ai, but few people should know that it has this kind of use. To put it simply, this is an automatic call function based on AI technology. After answering, Xiaoai will recognize what the other party said and reply with AI voice. After the call is over, it will also record the conversation process in text form. It is convenient for us to check later.
When you are chasing dramas, if you see your favorite British and American dramas but find that there are no translated subtitles, you will definitely miss it with regret.
Also in Xiaomi mobile phones, users can turn on "Xiaoai Translator" and find the "Real-time subtitles" function, so that bilingual subtitles will be generated in real time while watching videos.
real-time translation
Judging from the actual experience, its translation speed and accuracy are very high, which basically does not affect understanding. And not only watching dramas, but also performing well in online live broadcast, video conferencing and other scenarios, which can be said to be a representative application scenario of AI in the actual use of computers.
In fact, there are many other scenarios where AI can be used in mobile phones, such as AI-assisted signal enhancement. By detecting the user's hand usage habits through AI, the usage scenarios can be detected more accurately, so as to achieve a lower call drop rate and faster phone calls. Data transmission rate, wider network coverage and longer battery life; there are also intelligent voice assistants, photo sorting, AI video editing, face unlocking, etc., which meet the various needs of users in the process of using the machine.
3. Where does the AI on mobile phones come from?
However, AI is so powerful and practical, but it does not come out of thin air. The series of AI use cases we talked about all come from Mi 13, so where does the AI capability of Mi 13 come from?
This is about the second-generation Snapdragon 8 mobile platform that it is equipped with. The second-generation Snapdragon 8 mobile platform is equipped with a Hexagon processor specifically for AI computing. The processor is designed with an independent power supply system, and its architecture has also changed to better support the reasoning capability of the Transformer network on the terminal side.
Do you still remember the real-time subtitle function on Mi 13? The reason why it can present fast and accurate translation effects locally. It is also due to the improvement of the hardware performance of the Hexagon processor that the second-generation Snapdragon 8 mobile platform has become the first mobile platform in the industry to achieve INT4 precision. Compared with INT8 precision, INT4 brings 90% AI reasoning performance Improvement and 60% energy efficiency improvement, which enables the second-generation Snapdragon 8 to achieve a 4.35 times improvement in AI performance in natural language processing use cases.
In addition to the Hexagon processor, in the second-generation Snapdragon 8 mobile platform, the sensor hub capability has also been enhanced to double the AI performance. Generally speaking, the sensor hub mainly undertakes some low-power and light-load tasks, such as familiar voice assistants, visual wake-up, and picture classification.
In the second-generation Snapdragon 8, manufacturers can create richer application scenarios around the sensor hub, such as quickly scanning QR codes when the screen is off, or monitoring in real time whether there are strangers peeping at your phone screen , and close the message notification in time, etc.
In terms of imaging, the second-generation Snapdragon 8 mobile platform adds a physical connection called "Hexagon Direct Connection" between the Hexagon processor and the ISP, creating the first "cognitive ISP" in the mobile industry, and bringing "Real-time Semantic Segmentation" technology.
"Semantic Segmentation" can perform separate rendering processing on different areas and elements in the picture, such as distinguishing the subject and different backgrounds, or in portrait shooting, individually modifying the skin and facial features of the subject, its processing process It is real-time, and the speed is very fast. After the user presses the shutter, he can get the desired photo immediately without waiting.
at last:
As one of the hottest fields in recent years, AI has attracted the attention of many technology companies including Qualcomm. Today, Qualcomm has been at the forefront of the industry in the field of AI. A series of products represented by the Snapdragon mobile platform are constantly setting new records for mobile AI computing power. , PC, Internet of Things and other fields also have a broad layout. It is foreseeable that the terminal side AI will usher in a new round of great changes until it completely changes the user experience.
It is worth mentioning that Qualcomm officially released the white paper "Hybrid AI is the Future of AI" on June 16, clearly stating that hybrid AI with distributed processing in the cloud and terminals is the future of AI. In addition, Qualcomm also said that with the continuous reduction of powerful generative AI models and the continuous improvement of terminal-side processing capabilities, the potential of hybrid AI will further grow, and the prospects can be said to be very bright.
In general, whether it is from the actual experience or the white paper, Qualcomm does have a great advantage in AI on the terminal side, and it must be known that there are dozens of user terminals on the market equipped with Snapdragon and Qualcomm platforms all over the world. 100 million units, and hundreds of millions of terminal products pour in every year. It is also thanks to the support of Qualcomm AI technology that today's mobile phones, cars, XR and other smart devices have performed extraordinary. In time, with the continuous improvement of terminal AI capabilities, creating an AI world within reach seems to be It's not that far off either.