With the increasing popularity of generative AI, the ability of large language models has triggered more and more expectations for smart voice assistants.
We believe that the intelligent voice assistant supported by the ability of the AI large model will be launched soon. This prediction comes not only from the observation of the AI large model, but also from the understanding of Hongmeng. Hongmeng will definitely upgrade its large-scale model capabilities soon. Looking at the industry, it seems that only Hongmeng can achieve this in the short term.
The reason for saying this is that integrating the large language model into the terminal ecosystem is not only an algorithm problem, but also requires the AI model to understand, command, and interact with the massive capabilities in the operating system ecosystem. This is not something that can be done in the short term, but needs to continuously develop and adapt to AI from the operating system and basic software level, and finally let the new AI capabilities represented by the big language model and generative AI become the user experience in a natural way. a part of.
And Huawei is precisely the first manufacturer to bring AI capabilities to mobile phones. The Hongmeng system has been built and evolved based on intelligent technology from the beginning. It can be said that AI is one of the most colorful genes of Hongmeng. This gene will continue to grow to meet the imagination and expectations of more users for smart experience.
On August 2, Yu Chengdong, executive director of Huawei and CEO of Terminal BG, shared a video on Weibo. The content is to "spoiler" to users a major function of the HarmonyOS operating system: let Xiaoyi help users generate content.
Judging from the ability of Xiaoyixin displayed on Weibo, it already possesses the typical characteristics of AI large model and generative AI. An HDC invitation created by it has appropriate words, logical logic, and can accurately describe keywords.
This also means that the upcoming Hongmeng 4 is very likely to bring the industry's first AI large-scale voice assistant embedded in the system.
This seems to be a former alliance between wood and stone in "Red Mansion", unchanged and destined to happen. Hongmeng and AI will bloom in this summer.
AI gene
Hongmeng is consistent
As early as 2017, Huawei was developing AI capabilities on mobile phones. In the following years, a large number of AI capabilities such as AI photography and AI smart assistants have gradually become Huawei's signature. What many friends have not noticed is that the well-known Hongmeng nature today is also built on the foundation of AI technology.
Software decoupling, a new generation of Internet of Things technology and AI have enabled Hongmeng to have a more intelligent multi-terminal collaborative experience from the beginning of its birth. It integrates multiple physically isolated devices into a "super terminal" through soft bus and distributed technology. However, if "Hyper Terminal" wants to play a role, it must first solve two problems: one is that the operating system needs to understand the scene where the user is in, and the other is that the user can complete multi-terminal operations dexterously and conveniently. Both of these goals require a lot of AI technology to achieve.
In the HarmonyOS system officially released in 2019, you can see AI algorithms and AI experience everywhere.
Subsequently, in the rapid evolution of Hongmeng, intelligent capabilities have also evolved. In June 2021, Huawei officially released HarmonyOS 2, which brought a new experience of unified control center, super terminal, and universal card. A large number of users began to understand and get used to the new HarmonyOS.
In July 2022, HarmonyOS 3 will be upgraded, bringing six major upgrades including features such as super terminal, universal card, smooth performance, privacy and security, and the intelligent experience has also been further improved.
When it comes to the AI capabilities in the Hongmeng ecosystem, Xiaoyi has to be mentioned. In order to manage a large number of devices and services, the central role of the voice assistant is essential. Therefore, based on its AI capabilities, Xiaoyi has become a unified entrance for all devices in the Hongmeng ecosystem and a unified assistant for accurate service recommendations. At the same time, Xiaoyi is also constantly improving its own intelligent experience through technologies such as AI flexible deployment, optimized contact points, and collaborative services.
It can be seen from this that Hongmeng Ecology does not suddenly want to focus on AI, but that Hongmeng originated from AI, grew from AI, and has become a fertile ground for the continuous evolution of AI technology. The AI spiritual root planted deep in the soil of Hongmeng is waiting for its golden age.
Big model and little art
a natural fate
Today, smart voice assistants are ubiquitous, but there are still many problems with its experience. For example:
1. The AI dialogue is still not smooth, especially the multi-round dialogue and the poor medium and long-term memory dialogue experience, which leads to the lack of willingness of users to continue to open the voice assistant and put forward complex needs, which reduces the use value of the voice assistant over time.
2. The natural language understanding ability of the voice assistant is not strong, and the logical reasoning ability is insufficient. As a result, users must use strict vocabulary and pronunciation to accurately interact with voice, and a large number of daily and colloquial conversations cannot be recognized. This is very unfriendly to the elderly, children and users who do not understand AI voice interaction.
3. The voice assistant is only capable of invoking services, but cannot generate content. As a result, the value that voice assistants can bring to users is greatly reduced, and it is difficult to cultivate users' continuous voice interaction habits.
It is not difficult to find that the capabilities of large language models and generative AI can just fill in these shortcomings of voice assistants. But before, the developers of large models were generally concentrated in AI manufacturers, which were far away from the application scenarios of consumer terminals and the software and hardware systems.
How can we make up for this distance? The answer is that it needs a process that will happen when things are ripe and when things are ripe.
For large models to move towards terminal scenarios, the ability of the algorithm itself is first required. In this regard, Huawei has already accumulated enough. On July 6, the Pangu-Weather model from HUAWEI CLOUD was published in the regular issue of Nature. This research has successfully solved the key problems of existing AI weather forecasts such as the inability to deal with uneven 3D weather data, making it the first AI method with an accuracy higher than traditional numerical forecast methods, and the prediction speed has increased by more than 10,000 times, within seconds level to complete the global weather forecast.
After years of accumulation, the Pangu series of large models have reached the forefront of the world in many fields. Coupled with Huawei's accumulation in the field of AI development platform and AI development framework, the combination of large models and Hongmeng has become more smooth and comprehensive.
On the terminal side, Huawei has made extensive explorations of AI technology. For example, we have seen Xiaoyi's smart image search function before. This function is combined with AI multi-modal capabilities, and through the understanding of generalized semantics and model miniaturization processing technology, it is the first in the industry to realize mobile phone natural language image search.
From models to tools, from software capabilities to developer ecology, Hongmeng is ready to welcome the arrival of AI large models. To put it romantically, Hongmeng's AI cultivation for many years may be waiting for this summer.
The AI agreement of Hongmeng 4
it's blooming today
The possibility of large language models and conversational AI should bloom in the role of terminal voice assistants, and the flowering place may only be in the garden of Hongmeng.
From a speculative point of view, in Hongmeng 4, we may see Xiaoyi that deeply integrates the big language model. Although we still don't know its specific capabilities, but combined with the capabilities of the large model and Hongmeng's past performance, we can still predict what will happen next.
Here, we might as well have some imagination together:
1. Communicate with Xiaoyi like a real person.
The strong generalization and strong robustness brought by the large model greatly enhance the natural language processing ability of AI. To put it bluntly, it is to allow AI to better understand language, whether it is homonym, metaphor, or reference. This ability will undoubtedly be the focus of the next upgrade of Hongmeng 4. Next, we will communicate with Xiaoyi like a real person, even ignoring the true identity of its voice assistant.
2. See, hear, and serve.
It is a major feature of the Hongmeng system to wake up the service closed loop based on Xiaoyi. Then, based on the multi-modal understanding ability brought by the large model, there is reason to believe that Xiaoyi will strengthen the reach of services in the future. Let Xiaoyi listen and watch, both will become the service entrance of Hongmeng 4.
3. Keep simplicity to yourself and complexity to Xiaoyi.
When the big speech model was just emerging, many people were willing to deliberately say a long paragraph, and let AI understand and analyze the meaning layer by layer. If it was just a game at the time, then the logical reasoning of the large model in Hongmeng 4 will become an ability, and more and more complex tasks will be performed by Xiaoyi in the world of Hongmeng.
4. Not only looking for things, but also creating things.
For a long time, the role of voice assistants has been to find the entrance to services and information. But in the era of large models, voice assistants can not only be found, but generated and created. Next, perhaps we will rely on Xiaoyi to complete various tasks in office, life, and study, and help us create various content.
Let’s take these imaginations with us to see how many of them can be realized in Hongmeng 4. Hongmeng and AI, this alliance of wood and stone is destined to blossom and bear fruit this summer.
The answer lies in the next HDC. Of course, more love between Hongmeng and AI is still hidden in the future, hidden in the minds of thousands of developers, and hidden in the expectations of countless users.
