Every day we live and work in our country, we can not leave the center of mobile phone for every scene of traveling, consuming, socializing and entertaining. Every day, there are huge amounts of personal data and interactive information input into the mobile phone, and countless contents are output from the mobile phone, so people It is natural to start with the assumption that the phone is not able to use the data to do what? For example, to understand its user owner?
'Let the phone know you more' has always been the main direction of the smart phone era.When the artificial intelligence representative of the machine learning technology and the personification interaction features deployed on the phone, AI has become the main means of this direction.
With the unveiling of Kirin 970, Apple and Huawei have successively released mobile phones equipped with AI processors, and some brands such as Google and Samsung have also released their AI plans for mobile phones. Perhaps it is time to look back and forth and understand the user's perspective. What kind of answer sheet.How do we understand this need, whether there are some problems that can not be solved?
In order to understand how ordinary users view the 'mobile phone to understand you', we first conducted a small-scale survey, the results obtained is still very interesting.
Horror Valley effect: about the "cell phone understand you" human tangled
In order to make the questionnaire representative, we randomly selected 25 respondents from all walks of life in 2 artificial intelligence communities and 2 mobile phone communities to form a 100-person surveyed population. Respondents Basic understanding of the topic of mobile phones and artificial intelligence.
Our survey includes four questions, two judgment questions and two multiple choice questions.
First of all, our question is, do you expect to understand your mobile phone? As expected, most people choose 'yes'.
But then we ask if we are willing to allow the phone to read and understand all your personal data with security, including purchase behavior, chat history, travel information, etc. As a result, more than 70 people said they did not want to, .
In the next multiple-choice questions, we first ask the respondents to answer the question as to how well a mobile phone can understand you more. Options include: 1 Read your hobbies, 2 Understand your phone habits, 3 Read You (Including iris, fingerprints, exercise data, body function data, etc.) More than half of people chose 1, that is, read the user's hobbies.
Finally, our question is which of the following is what you need most: 1 smart secretary, 2 smart power saving, 3 smart hardware cleanup, 4 smart ecommerce recommendations. The results are somewhat unexpected, the most votes are the second smart Energy-saving, and smart e-commerce recommended is the lowest votes.
From this result, it seems that we can feel a bit tangled: We hope that the phone more understand themselves, but do not want to put their own information to the phone.We hope that the phone can understand their own hobbies, but the real expectation of the function is to individuals Information needs the least 'power saving'.
The subsequent communication with the respondents also confirmed our view that most of the respondents believe that cell phones understand themselves, of course, is good, but the phone to understand their privacy information or there is a big security concern. A function, the smart secretary, many people think that there is no practical use, e-commerce recommendations are likely to join the interests of the business-driven, but it is to save power and clean up memory, really quite necessary.
This may be a typical paradox: on the one hand we think cell phones are more intelligent, we can understand better that we must be right, but on the other hand, the insecurities that open up all private information, The uncomfortableness of being monitored by the phone prevents us from really accepting 'love and understanding' from the phone's AI, which may be the horrifying psychology of humans when facing cell phones.
Important Non-Critical Information: Mobile AI Understand How Users Start?
So, mobile phone users to read is completely a pseudo-proposition it?
Perhaps not necessarily, after all, there is a long edge between the privacy concerns of users and the upgrading of experience with machine learning - perhaps the best stage for AI to release its learning abilities on cell phones.
Here we first need to understand a question: what is the data we enter into the phone every day?
The first type of data is key data, including key biological data (fingerprint, iris, faceID, etc.), key data (account, password, phone number, ID numbers, etc.), behavioral critical data (purchase information, social information, travel information, etc.) and the other is non-critical data including audio data (telephone, voice input during voice), environmental data Heat, air quality, etc.), human-computer interaction data (user clicks and swipes).
And we really worry about leaks and prone to discomfort, the basic is the key data is collected and utilized, but the real phone every day to collect the most non-critical data.This part of the data for the user is basically meaningless, but may let The AI system on the handset learns from it, understands many user-related information, and feeds back personalized solutions through algorithms.
Based on this, the phone can understand your tangle may be from the application of non-critical data to crack.
First of all to solve the user interaction with the mobile AI from 0 to 1, have everything behind it.
Mate10 unique understanding of your posture
Of course, this is not to say that it is not advisable to understand and learn some key user data in mobile phones. In fact, applications and software will learn these data even if the mobile phone does not do so. On the contrary, the integrated mobile phone may bring more imagination.
However, most users today still have concerns about whether mobile phones can fully learn their own information.It is very difficult for such functions to be deployed on a large scale on the flagship aircraft which must be recognized by the market, so as to avoid unnecessary public opinion disputes and market conflicts.
Previously used as an experimental product of forward-looking technology, glory magic has been equipped with active learning user data artificial intelligence system, so that the system has the ability to perceive, analysis and proactive service capabilities, but even as a cutting-edge products, glory magic still let users Choose whether to open this ability, as well as layers of protective measures can be seen mobile phone users understand the proposition really hard landing.
The newly launched iPhone X, even as a full-fledged Apple product, has a single AI processing unit and a large number of AI features, but still choose to understand the user's habit of this level is not deployed.Is the AI technology focused on the camera and Camera and recognition unit.
Compared to mature AI solutions such as voice interaction and machine vision, understanding and learning user habits may be a hard-lovable, but also a hidden risk of technical ports.Apple's attitude is basically circumvented, but fortunately, Huawei does not .
We noticed that Huawei Mate10 just launched a series of discussions on mobile AI and did not choose to 'know the user around the handset' as a puzzle. "Without a radical plan to learn user core data, Mate10 seems to be at Non-critical user data understanding and personalized output under a lot of effort.
For example, the Huawei Mate10 is equipped with the Easy Talk function based on AI technology, which relies directly on the NPU processing unit of Kirin 970. The function is to learn the daily volume of voice calls, timbre, pitch, etc. The AI algorithm to optimize personal voice Model, thereby increasing microphone sensitivity.
The most important thing about this function is that it achieves the purpose of denoising the AI by learning non-critical data such as voice of the user (after all, it is certainly not privacy), knowing that for a machine system, the user's voice and other voices , There is no difference between the ambient noise.If you want in the long-distance or noisy environment to pick up the user's voice, we must understand how the user's voice, the user's language habits, how to control the sound volume in order to eventually achieve Accurate denoising and crawling.This Mate10 introduced the driving scene call quality optimization features, should also use this technology.
Another important area of user data learning is the user's habits, which include the user's mobile phone usage frequency, wake-up time, software habits, etc. For the user, these habits do not involve any privacy, but the user For mobile phones, learning and understanding these habits and reverse output may affect the entire mobile phone system experience.
Through Kirin 970's AI processing capabilities to achieve the machine learning and understanding of human-computer interaction habits, released Mate10 AI capabilities in system optimization in the consumer experience, the biggest change is the first Android comes with Chen Redundant Caton problem is circumvented, followed by the user habits to intelligently adjust the software frequency, reducing power consumption, disguised access to a lasting standby experience.
By mining these non-private but important user information, and taking note of the seemingly unrelated and fleeting data, the Mate10 is the first in a new era of mobile-phone literacy for the AI-era.
However, this is only the first step. The new postures on several key features are not the full picture of the mobile AI ecosystem Mate10 may open.
Rehearsal of the future: gradually plastic mobile AI ecological next step?
When electricity is produced, it is not marveled at its value, but at the very first to question the danger it poses. Until the exchange produces a man-made light, mankind is completely conquered. After that, electricity is still dangerous, but reasonable protection And safety common sense makes it into every household.
Today's mobile AI, is likely to go through such a process.I want to make the machine more people, better human services, we will one day be fully open human-computer interaction, let AI everywhere, but there must be enough Great value to continue to persuade humans to Caixing.
The way to persuade is nothing more than two clues: the undoubted security and the increasing application of imagination.
The former reaction in iPhoneX, Huawei Mate10 these AI phones, the most significant point is that the mobile AI chip brings localized AI processing power.Only effective cloud collaboration with the cloud, the data completely separate, personal privacy can not spread, in order to allow Users believe that the safe premise of man-machine data sharing.
And more importantly, the application of imagination continues to expand, through the calculation of AI processing unit support, coupled with Mate10 such adaptation models in perception, learning, understanding of the integration of performance, to build a very good development platform. AI's future will always be in the developer's imagination, AI clever way to read the user needs to be further improved.
Kirin 970 with the release of HiAI artificial intelligence development framework is reserved for developers stage. Coupled with Mate10's own function demonstration, leveraging the user habits through the application may be the main battlefield in the future.
Mobile AI proposition is a castle that can not be achieved in one step and must be gradually sculpted, while the hardware foundation of Kirin 970 and the platform preparation of HiAI have made some rehearsals for the farther future. Under the open development structure and the security of localized AI , Mobile phone users understand the next most likely to reach some vertical areas to expand.
For example, in the field of health, users selectively share health information, travel information and medical history, and the system will be able to learn user health habits by themselves and give smarter meals, exercise, nutrition habits, and care and rehabilitation processes. Very tempting.
For another example, in the area of business applications, mobile AI can play a crucial role as an industry assistant by learning about the user's itinerary, scene changes, and work habits, which may not be necessary for the average user but for business people Has a great attraction.
Another possible way to understand the user's mobile phone comes from the synergies between the mobile phone AI and other hardware such as handsets, watches, VR devices and mobile phones after the perception of integration, the mobile phone as a processing center to achieve collaborative data of mobile phones and feedback through Terminal computing power to establish a hardware system of AI closed loop, can bring a sharp rise in the application of imagination.
Discussed so much, just hope to help you more three-dimensional understanding of the phone AI in the end how to understand the user.Development of this process may be different with our imagination, but by no means meaningless.Human-computer interaction is always people-oriented, machine know of course Than human understand the machine more natural.
Let the phone understand you, is not impossible, but difficult, but fortunately challengers such as Mate10 is always there.