Intelligent assistants have entered human life for ten years. On a barrage website, I like to watch users "mocking" various smart assistants, asking them some weird questions, and asking them to answer the phone for me. Every time I can't help laughing. At the same time, seeing everyone complaining that the intelligent assistant is still "artificial mental retardation", it is inevitable that there is a trace of regret.

With the continuous advancement of digital technology, are there new opportunities for intelligent assistants and their industries, and can they usher in further development opportunities? At this year's OPPO Developer Conference, we saw more possibilities presented by Xiaobu Assistant.

Based on the technical capabilities of Andeverse, the "integrated digital brain with cloud and device", Assistant Xiaobu created a digital parallel world of virtual and real symbiosis at the Metaspace conference, and released the 4.0 annual version, integrating Xiaobu Space and many others. Innovative product functions allow users to experience the digital experience integrating virtual and real in advance.

(Liu Haifeng, President of OPPO's Digital Intelligence Engineering Division, made his debut at the Xiaobuyuan Space Conference)

From this, we can read about OPPO’s technological breakthroughs in the fields of artificial intelligence and virtual sapiens. At the same time, we can also grasp the pulse of mobile ecology and mobile interaction in the new context. When users become familiar with and love smart assistants, rely on its To complete more life experiences, Xiaobu, like a ship, is becoming a vehicle for more people to go to the new world of Shuzhi.

Xiaobu has never stopped evolving, and it also represents the continuous exploration of intelligent assistants, which are inextricably linked to each of us. Let's start from Xiaobu's evolutionary roadmap and have a glimpse of the other side of Homo sapiens.

Crossing the Seas: Rising Sea Levels

First of all, we need to clarify why the evolution of intelligent assistants is an important reference point for understanding the future of digital intelligence. Looking back at the history of machine intelligence, as early as the last century, robotics expert Hans Moravec drew a "topographic map of human capabilities", with the middle and low ground representing "arithmetic" and "rote memorization", and the hills representing "theorem" Proof" and "playing chess", the towering mountains represent "movement", "hand-eye coordination" and "social interaction". With the continuous advancement of human beings on machine intelligence, the sea level is also gradually rising, and the human capabilities at the low level are no longer unique. Gradually, some higher-level capabilities can also be completed by intelligent life. For example, with the development and maturity of pre-training technology, machine dialogue has even reached a human-like level in some scenarios.

Technology continues to expand the capabilities of artificial intelligence, and the intelligence of intelligent assistants has also made great progress in recent years, which also means that intelligent life forms will play an increasingly important role in our lives, familiar with and mastering digital life. It is no longer an option, but a necessary life skill. So Hans Moravec proposed: We should build an ark and adapt to seafaring life as soon as possible!

As an interactive portal connecting the physical world and the digital world, intelligent assistants are very suitable to become the digital and intelligent ship that the general public can take.

Building a boat: The evolution of Xiaobu's body and mind under the support of AI

To cross the sea of ​​times, we must first look at how the ship Xiaobu Assistant 4.0 was built and what kind of capabilities it has.

The five newly upgraded capabilities of Xiaobu 4.0, from active intelligence to emotional interaction, to digital intelligence multimodality, smart new experience and multi-device collaboration. From a technical point of view, we can summarize it into three aspects, which constitute the core of the living body of Xiaobu Assistant 4.0.

Soul: Hans Moravec believes that social interaction, emotional interaction, etc. are one of the most advanced human capabilities, and also the unsubmerged peaks in the "topographic map of human capabilities". As the concentrated expression of OPPO AI application, the new version 4.0 is constantly reaching these peaks, showing a more intelligent side.

Wisdom is an abstract and ethereal thing. In order for a machine to exhibit human-like intelligence, it should theoretically complete its evolution from three perspectives: First, memory. Humans can incorporate a lot of long-term state information into the brain’s algorithm. It will be called one day in the future, and stable long-term memory is a major challenge of machine intelligence, which is manifested in intelligent assistants, that is, it is difficult to communicate smoothly and naturally in multiple rounds of dialogue; the second is computing, the parallel computing capability of the human brain Very strong and good at solving complex tasks like analysis, decision making. With the rapid development of algorithms and computing power, AI has also begun to demonstrate human-like capabilities in decision-making intelligence, allowing intelligent assistants to evolve from "command passive response" to "intimate active service"; the third is learning, IBM's When Deep Blue defeated chess champion Garry Kasparov in 1997, its biggest advantage was memory and computing power. By 2016, when AlphaGo defeated Lee Sedol, deep learning made a substantial leap in machine strength. With the ability to learn, intelligent assistants can self-evolve and iterate to solve the problems of stylized interaction and low human-likeness.

Specifically in the 4.0 version of Xiaobu Assistant, we can already see that Xiaobu is bringing real and perceptible experience changes in the three levels of memory, calculation, and learning.

In terms of memory, proper memory determines whether the interactive experience of the intelligent assistant is natural and smooth. For humans, it is almost an instinct to make an immediate response based on previous information, and there is no need to mobilize the memory module at all, but such a simple thing is very difficult for an AI agent. Before the emergence of long short-term memory neural network (LSTM), the traditional neural network had no memory function and could not process long sequence data. In short, it could not remember the information of long-distance data, which was manifested in the intelligent assistant, and the user finished talking to it. "Talk to me for a while after dinner", and it will also ask "Have you eaten?" To avoid the situation where the user says "Qianmenlouzi" and the AI ​​says "crotch axis", it is necessary for the intelligent assistant to understand the context well, so as to generate a more natural and emotional expression. It is inseparable from the strong memory ability. Therefore, the researchers specially developed the memory ability for Xiaobu, so that the AI ​​can understand some key data at a longer distance in the chat process, so as not to forget it after learning it, so as to generate interesting and useful chat content, and users do not have to repeatedly emphasize some What has been said, the human-machine communication will be more relaxed and pleasant. With memory, intelligent assistants have a personified basis for sustainable growth.

In terms of computing, with the support of end-to-end computing power, Xiaobu Assistant can be equipped with more powerful algorithm applications. Based on the self-developed emotion recognition algorithm, Xiaobu Assistant has a single round of intelligence, skill guidance, multiple topic rounds, and emotion perception. Waiting for basic capabilities, and then introducing more cutting-edge pre-training technology, through the large model of 100 million to 1 billion parameters, to improve the generalization ability of language understanding, and alleviate the "mental retardation" problem caused by "intent understanding is not in place". At the same time, it pays attention to the combination of AI and knowledge computing, and builds a high-quality knowledge map, with a scale of 100 million entities and tens of billions of relationships, so that Xiaobu has enough knowledge reserves to answer all kinds of questions from users. The question and answer aspect brought a 2% to 4% increase in the effect. With some technical polishing, Assistant Xiaobu's understanding of colloquial expressions, analysis of user intentions, and warm emotional interactions have been effectively improved.

Not only that, Xiaobu Assistant can also accurately judge user needs, combine contextual scenarios, and then proactively provide services. Connect the various "breakpoints" of digital services to make the service appear coherent. What users feel is the smoothness and smoothness that conforms to the inertia of behavior. For example, after sensing the user's travel needs, Xiaobu will provide the service before the user travels. Basic reminders such as luggage preparation, hotel reservations, traffic conditions, and epidemic prevention policies, so that you can fully prepare for the journey in advance.

In terms of learning, Xiaobu's more intelligent side is also reflected in the ability to continuously learn and develop and evolve. Internally, Xiaobu will continue to learn and evolve according to the user's behavioral feedback, and become more and more "understood you": if he finds that the user's needs are not met, he will repeatedly modify his words and try again; after receiving praise from users, It will also be further optimized according to the word-of-mouth satisfaction system and continue to improve itself. Externally, Xiaobu will continue to learn the data resources brought by multi-scenario and cross-terminal, and continuously expand its capability boundaries and service scenarios. For example, through scene understanding, it supports multiple commands in one sentence, making routine services such as checking the weather, setting alarm clocks, and making calls more convenient and efficient. Just say a "open health code" command to Xiaobu, which can save a series of tedious manual operations. process. In addition, Xiaobu can support the control of OPPO's own devices such as mobile phones, watches, TVs, and Pads, as well as third-party brands of Xiaobu's ecology, and execute commands across terminals and devices, breaking data barriers and allowing users to enjoy full-scene services. The seamless switching is silky and smooth; it can also provide multi-dimensional data nutrients for Xiaobu Assistant's self-learning through the multi-terminal data synergistic feedback algorithm, further improving Xiaobu's intelligence level and service ability.

Following the road map of memory, calculation, and learning, Xiaobu is climbing step by step to the top of the "topographic map of human ability".

Body: For a long time, smart assistants have been like the heroine in the sci-fi movie "HER", with only voice but no body. In recent years, with the advancement of digital intelligence technology, some smart assistants have begun to integrate voice interaction and natural language understanding. , image recognition and other AI capabilities, the appearance and image have become more vivid, such as Microsoft Xiaobing, OPPO Xiaobu, Tencent Cloud Xiaowei, etc. This year's Xiaobu is also further iterated with technical support.

On the one hand, Xiaobu 4.0 has carried out a new upgrade at the interactive level, which supports users to interact with Sapiens in real time through multi-touch on the mobile phone screen, such as poking Xiaobu's stomach, touching Xiaobu's head and buttocks, Xiaobushu Homo sapiens can give corresponding feedback. The 3D chat function of Xiaobushu Homo sapiens launched in version 4.0 of Xiaobu takes it a step further. Through 3D scenes, AI-driven Sapiens and story settings, it supports multi-modal interaction and creates a real and natural chat. Scenarios, identify user emotions, and allow users to gain a new immersive chat experience based on chat interaction and game entertainment.

On the other hand, Xiaobu Space provides an interactive field for Sapiens, which is more immersive and interactive. Although XR equipment has not yet been widely used, OPPO has created a meta-space concept product based on the Xiaobu Assistant APP, allowing users to experience the charm of virtual and real integration under the mobile phone interface. Xiaobu Space supports users to create their own images in it, 3D visual effects and real character settings make digital life more immersive and realistic, using "second avatars" to socially interact with Homo sapiens and real people in the square to unlock more innovations How to play, such as going to the exhibition hall to watch the live broadcast of the conference, completing the online conference without leaving home, etc., and experiencing "The Sims" in advance.

Physicist Max Tegmark proposed that version 1.0 of life, its hardware and software, was obtained by evolution and cannot be changed. In the stage of life 3.0 represented by artificial intelligence, life can not only design software (culture) by itself, but also design hardware (body) by itself, from carbon-based to silicon-based. Obviously, the designable and moldable "body" image displayed by Homo sapiens is the inevitable process of the development of intelligent life to the 3.0 stage, and it also makes us feel the fun of interacting with silicon-based life in advance.

OPPO's concept of "science and technology for people" has promoted Xiaobu's assistant to develop and iterate in a smarter direction in body and mind, and become an intelligent life in a beautiful and intelligent life. With a solid physical and mental foundation, there is also the confidence to further explore the future of virtual reality integration.

Set sail: the technological side of the digital world

Consolidating the capabilities of intelligent assistants is only the first step. The second value point of this innovation lies in the exploration of the world of Homo sapiens.

Objectively and frankly, with the development of the mobile Internet for more than ten years, users' novelty of human-computer interaction has also greatly declined. Everyone is eager for new experiences, and new experiences will surely establish a new order in the mobile terminal market. At present, the technical direction is very obvious, that is, a term that has been repeatedly mentioned – the fusion of virtual and real.

At this OPPO Developers Conference, we can clearly see OPPO's judgment on the technical path and industry direction. The Xiaobuyuan Space Conference will create a communication and sharing space where the real world and the virtual world are intertwined and naturally integrated. Xiaobu 4.0 episodes Visual effects that sense and interact as one.

Along the route of Xiaobu's assistant and Sapiens, what kind of technology will they head to? From OPPO's move, we can see three dividends that are being released.

1. Technical bonus. The application scenarios of Sapiens continue to expand, but the technical threshold is still high. A Sapiens with high interactivity needs leading AI algorithms to generate and drive lip shapes, expressions, actions, etc., such as sentences generated by NLP algorithms It is necessary to precisely match the mouth shape so that users can have a visual sense of talking with a real person. In order to make the interaction not boring and fresh, it is only a few fixed actions that cannot go back and forth, and GAN generation algorithms are needed to participate in the construction and drive the actions of Sapiens. Sapiens needs to provide services in various scenarios such as banks, hospitals, schools, high-speed rail stations, etc. It is impossible for all companies to rely on their own research and development of basic capabilities. Through the Xiaobu Sapiens platform and OPPO's open ecological cooperation, it is possible to avoid using the underlying technology. Repeatedly building wheels can lower the technical threshold and accelerate the industrialization process of Homo sapiens.

2. Industrial dividends. With the wide acceptance of multi-modal human-computer interaction, there are more and more industrial demands for Homo sapiens, but the presentation of personalized appearance and skills requires art design, 3D modeling, skeleton binding, texture pinching, etc. A series of operations, the high production threshold hinders the large-scale landing of Sapiens. At present, Xiaobu has also accumulated corresponding capabilities on the Sapiens platform, providing personalized, high-performance, multi-scenario Sapiens services, reducing the threshold for landing applications, and helping Sapiens penetrate into more scenarios in the B-end market. .

3. Ecological dividends. The rich and prosperous Sapiens applications and services are inseparable from the innovative wisdom of individual developers and enterprise developers, so that developers’ creativity and energy can be quickly transformed into commercial returns. OPPO’s comprehensive layout and ecological construction in the AIoT field provides abundant opportunities. Wo's achievements are transformed into soil. As mentioned earlier, Xiaobu Assistant supports the control of OPPO's own devices such as mobile phones, watches, TVs, and Pads, as well as third-party brands, covering all types of hardware, which means that related applications and services can be deployed on multiple terminals, Covering users in the OPPO ecosystem, developers use OPPO to gain commercial value, and further attract more people to build a digital and intelligent world that integrates virtual and real, and the OPPO innovation ecosystem has entered a virtuous circle.

It is not difficult to see that with the continuous release of the potential of Xiaobu Assistant, in the future, it will not only play an important role as an interactive portal in the OPPO ecosystem where everything is integrated, but also spread the value of intelligent life in the entire mobile ecosystem, becoming More business and user interfaces with the digital world. Under the general trend of integration of all things and the integration of virtual and real, the existence value of Xiaobu is showing unprecedentedly.

Max Tegmark of the Future of Life Institute believes that the future of life with artificial intelligence is the most important conversation of our time. Assistant Xiaobu is taking us to participate in a warm and interesting dialogue with AI life. There is reason to believe that people born in this era of great development of artificial intelligence should work with intelligent assistants to achieve better each other.

A wonderful journey is waiting for us to sail out to sea. Let's start with the phrase "Xiaobu Xiaobu".


Leave a Reply

Your email address will not be published.