Mobile AI Development Ecology Scramble | Mobile AI Travel Map < three >

Let us think about the distant past, what is the reason why we abandon the same function machine, the choice of smart phone?
Is it because of the beauty value? Interactive freshness? I believe that the vast majority of users, because the app mode brings too much practical value, people around the use of, and they can not even follow up. So amid the greatness of the Lord is not only to subvert the shape of the phone, but more importantly to the future of the mobile phone ecosystem opened the entrance. Until today its energy and imagination are far from exhausted.
By the time the AI era begins, the logic still seems to be universal. After all, AI algorithm is characterized by strange things, do anything.

Some people use AI to do medical care, some people use AI as customer service, and even some people use AI to predict the time of death, how to put these amazing ideas, all transplanted to the mobile phone, it is clearly the era of real big business.
But for the new artificial intelligence world, it is not easy to bring the developer and the development ecology into their own platform, thus forming the industrial barrier under the proposition of the mobile phone AI. The battle for the development of the ecology, in fact, has been playing a big game between the Giants.
Millions of smart minds are worth more to AI.
Look at this year's centralized release of the many true and false "Ai phone", will find a lot of interesting phenomenon. For example, last year, Mate10 first released the scene recognition, data label photo mode, the results of this year the brand's flagship machine all joined a similar function.
Of course not that this function is not good, indeed scene recognition + shooting can solve a lot of problems, bring experience upgrade. But the question is, is it too similar? How can AI, with so many abilities, become the same "student head" at the end?
In fact, the current high-completion AI solution on the mobile phone, there are image recognition, environmental understanding, image enhancement, NLP, speech processing, such as several major categories of dozens of kinds of capabilities. At a minimum, you can improve the performance and experience of the six mainstream applications, such as live and short video, photography, social, shopping, AR, and translation. The possibility of creating unknown and popular applications is even more enticing. After all, mobile phone manufacturers themselves can never do many AI applications, the real can let the general public accept mobile AI, is millions of smart minds, and even a new business model based on AI solutions.

But the full technical ideal wants to get into the bone sense of the application reality, there will always be a little bit of distance.
For example, the lack of AI-specific processing power in mobile phones, many AI tasks on the mobile phone experience is very poor, or simply can not run up. There is a lack of platform and API open premise, developers do not know how to transplant AI model to the mobile phone. Farther away, developers are not daring to engage in the new development of mobile AI in the context of uncertain future benefits and business value.
Phone AI itself to the application of the diversity of calls, and developers lack of technical support and business assurance, dare not rushed into the mobile phone AI development, almost become the main contradiction of the mobile phone AI topic. If not, we may face the embarrassment of seeing only two or three new AI applications per year.
Of course, contradiction is always an opportunity. Especially for the players mastering the technology-initiator advantage and ecological aggregation, it is probably the best layout for the future when the industry is unable to accomplish one thing in general. Now the battle for the development of mobile AI applications has been quietly being waged among a handful of giants.
Hiai and Tflite: The Giants ' ecological scramble has already begun
based on the terminal AI acceleration capabilities provided by the Kylin 970, as well as  and Glory have launched three of the flagship products equipped with AI chips, the role of Hiai architecture is to open up the developer link, the introduction of ecological development capabilities.
So far, the Hiai architecture has been upgraded to version 2.0, almost all of the mainstream deep learning development frameworks, and has launched a development course and two XXX hair motherboards.

Through open chip capabilities, algorithmic capabilities, and application capabilities, the entire Hiai architecture is now available to developers with a relatively complete range of five engines and full interfaces. This gives developers the ability to target AI-capable platforms, avoids the technical difficulty of collecting data and training the process from the start, and a lot of time and money costs.
So far, we have been familiar with the quick, vibrato, beauty, and a lot of shopping, social applications, have disclosed the cooperation with  and Hiai architecture. For example, a racer will develop new live effects, gesture and limb recognition, scene recognition applications based on the Hiai architecture, and will also develop new compression models to use AI effects in a weak network environment. The
is similar to  approach to implementing AI at the terminal, but it is quite different, presumably for TensorFlow Lite, which Google opened at the end of last year. Unlike the Hiai based on  AI chips and products, Tflite is essentially based on the deep learning development framework TensorFlow, but is designed to help developers develop and run learning models on local devices. This also leads to the feature being biased to one end of the algorithm development, rather than the application and commercial aspects.

But according to reports, some of the applications in Pixel 2 have been developed based on Tflite, and the future of a large probability of the emergence of Google mobile AI chip, will also be combined with tflite. At present, many great gods on the TensorFlow have made a lot of playful application models based on Tflite. These applications are far from Chinese users, but many ideas are valuable for domestic developers.
It is worth mentioning that, although Apple has launched the AI chip A11, but has not launched the overall support of AI eco-development platform-based products. But that doesn't mean Apple doesn't value the developer segment. Last June, Apple opened its own machine learning capability called core ML on the iOS Development Board, and now has the Vision API and the Natual Language API two API interfaces that enable developers to develop machine vision and natural language processing capabilities.
for the time being, Apple prefers small-scale, low-level capabilities to enable developers to improve the app experience in an iOS environment, rather than making disruptive development. According to Apple's characteristics, he tends to prefer to hoard the technology after maturity, with superb engineering ability for disposable release.
In any case, the Giants ' scramble for the AI development Ecology has set the course. At present, the characteristics of fighting each other can be seen as giants based on their own technical strength and strategic demand for water test, which led to a lack of uniform standards, but also to the developers to bring savage growth opportunities. In any case, the core of the battle for the future is long established.
What is the Battle of the mobile AI?
In the context of the new technology to scour demand and cognition, the platform and the various fields of developers want to integrate into the ecology, is always determined by a number of factors.
The first priority is to reduce the entry threshold.
This threshold includes many aspects, such as technical access threshold, trial cost, learning finished product, migration cost, and compatible cost. Developers can not spend too much time and money to try the unknown market proposition, and do not want to because of platform compatibility, the framework of migration and other issues to make their own. Even developers who do not understand the algorithm, want to enter this field to contribute wisdom and traffic. And these are the responsibility of the platform side.

Of course, this is not to say that developers should not understand the algorithm. Instead, different developers have different needs and values, and the platform has to be selective, allowing developers to find the most appropriate posture for their release.
Secondly, good ecological ability is inseparable from the good energy-empowerment scheme. such as the profit-splitting policy, market guidance and so on. For developers, especially in highly competitive markets such as China, business value is always a prerequisite. platform in the provision of good technical solutions, must consider and effectively guide developers to reasonable access to traffic and business returns.
Finally, in the case of weaker market cognition. A sensational "big strokes" case may be more convincing than all technical parameters and market analysis. Once the platform has hatched a singular point application, the market will soon see the business value that is growing based on the AI chip and AI Mobile development architecture. Thus confirming the reasonable way to enter the AI age of the mobile phone. Logically we all know what the future will be, but the real timeline has to be a case to open up the future.
Overall, mobile AI eco-development is like a castle built in the future. We know that flowering is the truth of AI, and that the Giants that hold the technological edge have begun to break ground. Perhaps a little bit of time will be the last seasoning needed for the entire AI mobile development.
For today's Chinese developers, it is clear that the Hiai architecture is a better choice, and for developers with a high level of technology, it may be better to develop their own development system by being able to integrate different development scenarios.
Perhaps an example of the AI development ecosystem for the value of mobile phones: Today, a smart phone to say the best, the result is not equipped, it is basically waste plastic.
In the same way, the value of AI is to be able to quickly pick up an unknown AI application that is not far away. In this way, the strategic barrier of mobile phone AI is not at all in the marketing word "AI", but in the face of the development of ecological technology entry threshold and ecological construction plan.
As far as we can see, this is still a giant technology game. Of course, everything is still full of variables.

Mobile AI Development Ecology Scramble | Mobile AI Travel Map < three >