Shulou(Shulou.com)11/24 Report--

Recently, iFLYTEK held the 1024 Global developers' Festival in 2023, and the brand-new iFlytek Input version 13.0 was unveiled, which is also approaching the 13th anniversary of the birth of iFlytek Input. The biggest highlight of this big version upgrade is the launch of the industry's first generative AI input method.

Nowadays, the development of generative AI is in the ascendant, and it is an inevitable step to combine it with the input method of the basic entrance in the era of Wisdom. In August this year, iFlytek Input first released an AI creation assistant based on the big model, covering six scenes, adapting to more than 60 applications, and pressing the "start button" of the big model application. This time iFlytek Input 13.0 is based on the brand-new cognitive big model of the input method, integrates the generative AI ability matrix into the main scene of the input method, and presses the "fast forward key" of the large model application again.

Its vision is to build a new model of human-computer interaction, to understand the needs of users in a second, to generate wonderful expressions, and to connect multiple services.

What does it mean behind this vision? How did iFlytek Input achieve it? It all starts with the "input method" itself.

The "third stage" of input method is coming. Input method is not only the basic tool of human-computer interaction, but also the entry-level product of the Internet and the era of the Alliance of everything. Over the years, the input method has been developing mainly with the progress of human-computer interaction technology and the enrichment of intelligent terminal products, but on the whole, its development can be divided into two stages.

Take Chinese input as an example, the first stage is five strokes / pinyin / handwriting input, that is, we are familiar with keyboard typing and handwriting recognition. In the second stage, with the progress of AI speech recognition technology, speech input has gradually become a more and more important input method.

According to a survey conducted by Analysys at the end of 2022, the penetration rate of voice input in China has exceeded 40%, making it one of the most widely used input methods for users.

It is worth mentioning that it was iFlytek Input who pioneered Chinese voice input. Thirteen years ago, they innovated through this source and opened a new chapter in human-computer interaction. Over the past 13 years, AI speech input technology has continued to evolve, and great progress has been made in recognition accuracy and supported language types. In the case of iFlytek Input, the current speech recognition rate has exceeded 98%, and has supported speech input in 26 dialects, 5 national languages and more than 30 foreign languages, realizing the transformation and upgrading from tool products to intelligent interactive products. no, no, no.

At the same time, iFlytek Input provides users with unique input efficiency value by virtue of continuous and innovative voice input. According to the latest set of data, iFlytek Input's daily voice penetration increased by 18%, daily voice calls increased by 35%, and the number of daily active users increased by 10%, thanks to iFLYTEK AI voice technology, which provides strong support.

Technological progress is only one of the factors that promote the development of input mode. Input method not only goes deep into our interconnected life, but also deconstructs our use scenarios to meet the ever-changing needs of users. At present, with the increasing maturity of 5G, AI and other technologies, the era of the Internet of everything is accelerating, new user needs are emerging, and the mission of human-computer interaction carried by the input method also needs to be changed.

So, what will the next generation input method look like after voice input?

Mob Research Institute recently pointed out in the industry insight report of China's third-party input method in 2023 that "AI + input method + scenario" is becoming standard in the industry, and further they believe that "input method + AI big model" will become an important trend in the future.

One thing is certain: "AI" will certainly become the focus of the next stage of the development of input method. In fact, "speech input" itself is the embodiment of the development of AI enabling input method, because artificial intelligence itself is the most cutting-edge form of human-computer interaction, and input method, as a basic tool of human-computer interaction, is also deeply influenced by the evolution direction of AI technology.

From the perspective of AI, from the beginning of the popularity of ChatGPT, the generative AI cognitive model has set off a new round of technology wave all over the world. In the past, people always said, "AI changes the world", but the answer to how to change has always been vague. The development and application of generative AI seem to make people feel clearly for the first time that "AI can really change the world" and change the way we produce and live.

It can be predicted that artificial intelligence is ushering in a large model era of "wisdom emergence". In this era, AI large model will empower thousands of industries, reconstruct productivity from the bottom, serve as an input method for human-computer interaction, and deeply integrate with AI large model capabilities, which is naturally a necessary step.

Generative AI input method, this is the first specific and clear answer. After determining the direction and trend, the next question is: how on earth should generative AI be integrated with the input method?

Different products may give different solutions to this problem, but the answer given by iFlytek Input version 13.0 is: with the blessing of the customized cognitive model of the input method, iFlytek Input has the ability of intelligent expression and intelligent connection. and has the characteristics of self-adaptation and self-evolution.

As a new generation of input method, iFLYTEK first reconstructs the interactive interface in order to help users understand the digital world better.

It is understood that more than 10 years ago, iFlytek Input defined the basic form of the modern input method of the upper toolbar plus the lower keyboard.

With the deepening of social digitization, users' demand for input method goes beyond simple text input. Therefore, iFlytek Input is thinking about taking another important step forward in human-computer interaction.

The new design of "dynamic Horizon" is the digital window that iFlytek Input strives to create, leaving more space between the screen and the keyboard, giving the recommendation area richer attributes. CTOnews.com also made a pre-emptive experience of Xunfei input method 13.0. specifically, in iFlytek Input 13.0, open the AI input mode, you can open the "vitality horizon" above the keyboard.

IFLYTEK saw that under the wave of the new generation of AI, users' demand for input methods has been comprehensively upgraded. It is an irreversible trend from simply inputting text to actively generating expression.

Intelligent expression means that the input method should be able to understand the user's intention, speak for the user, better complete the expression for the user, and actively help the user to create content.

IFlytek Input 13.0 has a built-in cognitive model of input method that has just been released, which is endowed with carefully trained generative AI ability. Next, let's take a look at iFlytek Input 13.0 how to help us better complete the expression.

For example, in the chat scene, if you want to tell everyone not to drink in the dinner group in the evening, and you don't know how to express it, you can type "tell everyone I can't drink" in the high EQ communication in the dynamic window. IFlytek Input generates a high EQ appropriate expression based on your input, and you no longer have to think hard or search the Internet.

For example, if you need to write comments for restaurants on the group buying website to participate in discount activities, you can also use iFlytek Input's shopping evaluation AI service to let it create an evaluation copy according to your meaning, so as to avoid the embarrassment of not being able to write.

You can also ask iFlytek Input to help you polish the text. For example, if you write a sentence "the moon is very round tonight", and if you think it's not beautiful enough, you can use the AI service of "embellishment" to help you polish this sentence, and you can choose different styles, so you don't have to worry about not having enough literary talent anymore.

There are many similar use scenarios, all from the concrete needs and specific scenarios to solve the problems encountered in the actual communication, so that you have never thought that "input method can still be so useful."

Not only that, as a new generation of generative AI input method, iFlytek Input is breaking away from tool attributes and turning into an assistant service, providing complete services to the right people at the right time.

For example, it can intelligently match multiple services in the input scene based on the user input content in different scenarios, and it can also connect the content and services outside the input scene according to the user input intention.

For example, when you are ready to post a moments, open iFlytek Input 13.0, you will find that it has intelligently matched the AI service to the "moments" copywriter, similarly, when you want to post a small red book recommendation post, open iFlytek Input will also find that it has understood your intention, matched to the "recommended copywriter", so you can use it directly, the whole process is very natural.

At the same time, iFlytek Input 13.0 can also intelligently connect content and services outside the scene according to your input intention. For example, in the chat scene, the editor talked about a restaurant with his friends and typed "this store". The input method guessed my intention according to the editor's past chat history, and successively thought of "sincere" and "good". The activity window of the input method is directly connected to the service outside the chat scene of "help you expand", and the editor tries to use it. Sure enough, it helped me write a paragraph describing in detail where the store is good. The random experience of this input service across scenarios makes the whole process of communication easy and efficient.

In addition, self-adaptation and self-evolution, that is, in the process of your long-term use, iFlytek Input will know more and more about your input scene and expression habits, so that whether it is auxiliary expression or content creation, it can be more accurate in line with your intentions and preferences, that is, more and more understand you, so that the input method can also be thousands of people, and better meet the individual needs of each user.

This is iFLYTEK's answer to how the generative AI input method falls to the specific product. From the experience of CTOnews.com, it does achieve the wonderful integration of "AI" + "scene" + "input method". It can obviously improve the communication, expression efficiency and effect of users in various scenarios, over time, with the continuous improvement of generative AI technology and the product itself. It is believed that it can really realize the revolutionary vision of "building a new mode of human-computer interaction".

Spark big model, iFlytek Input vision behind the technology base from version 13.0 to look back at the 13 years of development of Xunfei input method, AI is their killer mace for 13 years, at the same time, with the needs of users to "keep pace with the times" continuous innovation, is the fundamental reason why they can become more and more excellent.

Cheng Kun, general manager of iFLYTEK input method business, once said in an interview with the media that when they made the first version of iFlytek Input, their original intention was to improve input efficiency and provide the ultimate service. The weapon to improve efficiency is "AI". Whether it is unique Chinese voice input or basic input experience, AI is used as the basic support.

When the demand for universal services is met, new requirements will arise, such as new requirements in the face of device diversity in the era of everything intelligence, new requirements in the face of the complexity and diversity of input scenarios, and new needs in the face of more diverse user groups. IFlytek Input has never stopped, constantly discovering new requirements and solving new problems. For example, in the last large version, iFlytek Input also realized the capabilities of noise, speech recognition in the far-field complex environment, non-switching of more than 20 dialects, and multi-scene coverage of AI intelligent expression assistant.

Because of this, iFlytek Input can continue to be recognized by the public. According to the latest data, in October 2023, under the background that the input method industry has matured, their products DAU, daily voice penetration, Japanese tone usage, and the percentage of users after 00 can still achieve a breakthrough, and the growth rate of user scale continues to be the highest in the industry.

This makes us believe that in the next era led by generative AI input method, iFlytek Input will bring more room for imagination and stronger competitiveness, so that the basic thing of "input" will burst out new vitality in the world of the Alliance of everything.

And the driving force that we can believe so much lies in the solid technical base behind iFlytek Input-what I mean here is naturally the iFLYTEK Spark AI cognitive model.

In early September, the iFLYTEK Spark Cognitive Model was officially opened to the whole people. You should know that iFLYTEK Spark was only officially released in May this year, and no more than four months after it was officially opened to the public, the speed of its technical iteration is already staggering.

At the same time, in the large model evaluation reports released by authoritative platforms such as the China Enterprise Development Research Center of Xinhua News Agency Research Institute and the Massachusetts Institute of Technology Review China, iFLYTEK Spark V2.0 has achieved the first overall score, and is regarded by MIT Technology Review as the "smartest" domestic model.

In this iFLYTEK Global 1024 developers' Day, iFLYTEK Spark also ushered in the upgrade of V3.0 as scheduled, the capabilities of the seven dimensions of general artificial intelligence were further upgraded, and the code ability of the large model, as an effective means of linking the "digital world", was heavily upgraded, surpassing ChatGPT as a whole, and tamping the strength of the first square of domestic application of large models.

In terms of technological innovation, we say to support this breakthrough in iFlytek Input, that is, in terms of multimodal capabilities, iFLYTEK Spark has realized the alignment of voice, image, video and other modes into a unified semantic space to achieve semantic penetration, cohesion and multimodal collaborative emergence. Therefore, we can use a series of unique innovative features on the underlying technologies of generative AI, such as "AI skin", "AI avatar", "AI word-making" and so on.

There are many similar leading technologies, such as the input method cognitive model mentioned earlier, which realizes the aggregation of multi-tasks such as recognition, translation, language classification and so on, so that the information can be shared. Therefore, in the input method to achieve a variety of dialects and languages do not switch automatic identification, translation, these are also leading the industry experience.

Finally, the spark cognitive model is building an industrial development ecosystem of "1mm N (major industries) + X (subdivided industries and products)", and iFlytek Input is an important industry application of "wisdom input". With the continuous improvement of the whole ecosystem, the tight connection between iFlytek Input and various industry application scenarios is bound to usher in a qualitative development, which also brings a broader imagination for iFlytek Input's development to the B side.

Conclusion since the emergence of mobile intelligent terminals, "how to input more comfortably and efficiently" on the touch screen is the most basic problem with the development of the industry, but many products have not been solved.

‎ but iFlytek Input didn't give up because it was "difficult" to do, and fortunately, it happens to be a natural fit for iFLYTEK, which specializes in AI technology.

Because in the process of human-computer interaction, the essence of "input" is to let people's intentions be understood by machines. From this point, keyboard typing and handwriting that we were once familiar with are the products of compromise due to the lack of machine intelligence, and AI is the ultimate answer to this matter.

IFlytek Input has anchored AI from the very beginning, and has persisted in strengthening AI's input method empowerment from all aspects of product positioning, function and design for many years. The latest iFlytek Input 13.0 is the forward-looking layout and space occupation of iFLYTEK on the inevitable trend of "generative AI input method" when the generative AI era is coming.

Today and in the future, I believe that under the guidance of iFlytek Input, "input" can become easier than ever before, and human-computer interaction can really usher in a change that has not been seen for a long time.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.