In March, all mobile phone brands started their offensive in the new year. Interestingly, last year's AI, which was a few of its own, has become a popular keyword in the domestic mobile phone circle this year.
Or mobile phones with or without an AI name will generally instill an understanding of the consumer: With AI, your mobile phone will be more capable. Of these, the most mentioned is of course the ability to photograph, and other things such as beauty , Translation, reading, is also a key promotional field for manufacturers.
Similar propaganda looks more, it seems that we will form a point of view: The role of AI is to enhance.
AI can universally enhance the capabilities of mobile phones and applications. Of course, this is true. But after this narrative strategy is used much more, we are likely to overlook another possibility: As a technology with a wide range of different types of solutions, AI can completely As a basic condition, let developers create a mobile application that we are not familiar with today or even have no conception. And just when this 'unknown ability' becomes popular, the value of AI in mobile phones will be irreplaceable. .
Where does this 'unknown' opportunity come from? What are the conditions for incubating it? Here are some of the brain caves built on known conditions regarding mobile AI.
Where are the opportunities?
The reason why we are looking forward to some new and popular application experience in the AI era is because all the popular mobile applications have been based on the computing and hardware architecture that was established at the beginning of the smart phone. Chip + touch screen interaction + Camera+GPS, which is the basic carrier of all social, O2O, shopping and entertainment functions in our mobile phones. The basic conditions for these capabilities are fixed early on and the developers are using these foundations to flash and maneuver to find ways to build new experiences and Business model comes out.
And AI is changing these foundations.
For now, the basic capabilities that AI brings to mobile phones can be mainly reflected in four aspects:
I. Video and live broadcast: Everyone knows that AI is changing the way we understand and process video. We use real-time computing through AI chips, face and gestures in mobile video, human pose recognition, and portrait segmentation, video stylization. Such operations are becoming possible.
Second, AR: Making better augmented reality experiences on mobile phones, unable to leave AI's help. Deep estimation, light estimation, environmental understanding, and other algorithmic values can help AR that is more in line with market expectations. And these capabilities are not only Acts on AR.
Third, photography: AI takes pictures and stylizes deeply. It is not new. The next step is to use AI for image enhancement. Handling light and video capture and dynamic capture will become new highlights.
IV. Recognition: Using AI to identify words and commodities, scenery, etc. is becoming a new hot spot. Capabilities in this area may bring new opportunities to many mobile phone applications. For example, photo translation, handwriting recognition and recognition, text emotion recognition, etc. They can be widely used in translation, input and social fields. Image classification, image super-resolution, object recognition, etc., can also be used in self-media, social networking and shopping.
With so many new capabilities and complex algorithms behind it, it is actually giving developers new 'ingredients'. The simplest way to use these 'ingredients' is to mix them with 'old dishes', such as adding AI to short videos. Special effects and AI recognition capabilities; more complex is to form a single 'AI new dish'; a deeper step is to re-understand the basic capabilities of mobile phones, to conceive of unknown functions that have not appeared before.
In general, the core change that AI brings to mobile phones is that mobile phones can understand the real world through cameras.
Moreover, this function will inevitably become stronger and stronger. How to use this to integrate functions, create new applications, and establish their own flow logic is the real opportunity that today's mobile AI and mobile developers are facing.
Who will provide opportunities?
For mobile phones, there is a good idea or a foundation for development. It is not a chicken-egg, egg-and-chicken problem. It's just like cooking. There must be ingredients before chefs can try to develop new flavors. Make new dishes.
For the proposition of mobile phone AI, an AI application that can not be tried by a consumer and can be heated up must be based on sufficient development.
Looking at today, this development basis must be based on the three potential energy bodies that form an orderly ecology:
First, the open AI chip: The industry currently develops AI architectures and applications, relying primarily on cloud computing to provide computing power. More back-end, is Nvidia and Google's cloud computing AI acceleration chip at work. But this model comes to the mobile phone side Obviously it is not a long-term solution. On the one hand, uploading user's videos, photos, and other data to the cloud has privacy and security issues. On the other hand, this process takes time, causes stabling and delays, and is invalid when there is no network.
Without AI chips, handsets that only deploy AI algorithms on specific hardware are faced with problems that cannot support the development of unknown AI applications. Because new applications will certainly require CPU, GPU, and AI special processing units for collaborative support. Missing handsets for common AI tasks With the support capability, users can only experience several AI functions that are set in advance and cannot meet the new AI experience. This is why Apple wants to develop the A11 bionic neuron chip, Huawei chip to develop the Kirin 970, and it needs to be developed based on chip construction. Ecology.
Second, platforms and frameworks enable AI development: Of course, only chips can't. Developers can't disassemble chips to see what they can do. Therefore, they can open AI capabilities at different levels and access mobile AI platforms with mainstream AI development frameworks. It has become crucial. This is like setting up a channel and market between computing power and developers. Developers need to know how to get into AI and know what benefits they can get from mobile AI.
For example, the HiAI architecture based on the Unicorn 970 chip can be regarded as a platform and developer layout for Huawei's AI development ecosystem. It is a mobile phone AI development platform to reduce the developer's entry cost and implement the benefit sharing and intellectual property protection plan. Based on the guarantee, after all, the ecological environment has been established, and the policy has given it a boost, so that 'investment promotion' can be activated.
Third, developers who are aware of opportunities: In the proposition of mobile phone AI, the platform is only black land, and the developer is the farmer who chooses how to sow and harvest. However, the first task that is placed in front of the developer community is to Recognizing that there is this fertilizer field, there is the possibility of AI development. Then it is time for the technology to break through and open the brain to think about creativity. It is equally important to establish an AI-based business model and industry barriers.
The virtuous circle between the above three can be seen as the basis of an agro-ecology. Of course, as consumers, we finally hope to see food on the table, but there is no orchard farm. Where does the food come from?
Mobile AI opportunities, not just in APP
Farther away, we should see that the AI brought to the mobile phone is not only the possibility of creating new explosive applications. If the technical conditions are further matured, we may get more from the mobile phone AI.
For example, we experienced the AI speaker battle last year, and this year we see more and more hardware, especially appliances are entering AI interaction mode.
But have you ever wondered that since the main value of AI speakers is interactive, not 'sound quality', why do you have to put a lot of hardware instead of directly talking to your mobile phone and let it help you realize content and shopping? Control the function of the home system?
After all, the mobile phone is a weapon that never leaves the body, and there are no restrictions such as wake-up distance.
This is because the AI function of the speaker and IoT equipment needs to rely on cloud computing to run on the phone is too slow, power consumption is too high.
These problems may be affected by technical defects such as AI chips and 5G, as well as the ecological development of mobile phone AI. If we can achieve efficient terminal AI calculation and fast transmission, then mobile phones can completely replace the speakers or other hardware carriers and become The only entrance to AI's life. This technical possibility will be the biggest surprise mobile phone brings to the AI industry.
Talk directly with mobile phones, use mobile phone sensors and cameras to understand and control the IoT system, achieve AI of business, family, entertainment scenarios, and even open the phone and driverless system, then the value of mobile phones will once again leap.
In the past, we have been speculating that the centralization status of mobile phones will not be eliminated in the AI era. But if AI happens on mobile phones, is this speculation not broken?
In short, AI brings to the mobile phone, certainly not only the enhancement of local capabilities, but also re-organized the basic elements of mobile phone application development. Under new conditions, AI or mobile phone can go?
This is an unknown number that makes people move.