Tmall Elf, Xiaodu, iFLYTEK and others went into battle.
AIGC terminal wave started?
Text | Intelligent Relativity (aixdlun)
Author | Ye Yuanfeng
It seems that this may be a pair of normal black-rimmed glasses, which you put on politely and politely;
In fact, it also contains a "small partner", you can communicate with it at any time, talk about things or ask various questions to accept its help; You become a geek of science and technology, and become a cool protagonist in a sci-fi movie who travels the world with a super-powerful "minion" …
This pair of glasses is an engineering version of smart glasses with a large model that was recently released by Tmall Elf.
According to official information, this new portable terminal demo has built-in AIGC capability, which can accompany users around the clock through continuous voice interaction.
Judging from the videos revealed by bloggers in bilibili, dialogue, recording, life skills … in the form of bone conduction, a large number of interactions have been easily realized without anyone noticing.
When you need to find a quality restaurant when shopping, just ask, and the answer will reach your mind, bring a puzzled friend and set off;
When the weight loss plan has not been successful, and you are confused about the calories of delicious food, you can immediately get the recommendation of appropriate food by your ear, without worrying about exceeding the calorie limit;
Lying on the sofa, blowing air conditioning and playing with the glory of the king, when you want to know how to play the auxiliary position, your friends have already arranged what heroes you should play and how to play …
Although artificial intelligence has been shouting for decades, the arrival of AIGC may be the first time that ordinary people can directly feel the surging power of technology.
In fact, in addition to Tmall Elf’s smart glasses, AIGC hardware has seen a wave of centralized emergence in the past.
Baidu fully accessed ERNIE Bot to a small extent, and even took the opportunity to launch smart phone products for the vertical field;
Once the Spark of Iflytek was released, it was announced to access its learning machine products and smart office products;
In addition, Huami launched the ChatGPT dial on its new Amazfit Falcon sports smart watch, and Xiaomi, which has many smart terminal products, broke out to develop a large model to access its ecological products.
Abroad, Apple approved its AppleWatch hardware products to access AIGC capabilities, and Innovative Eyewear, a smart glasses company, announced that it would launch the first smart glasses supporting ChatGPT.
The above-mentioned Tmall Elf smart glasses directly ranked first in the same category of Tmall platform on the day of sale.
All these indicate that the first generation of AIGC hardware is welcoming an outbreak, which also shows the development path and value of personalized and portable products under the vertical AIGC.
one
large-scale model
Want to be closer to "users"
ChatGPT, Tongyi Qianwen, ERNIE Bot, iFLYTEK Spark …
Large-scale model products have emerged in a concentrated way in the past, and each of them has a sense of existence, which is very lively.
In fact, these products that can chat and interact are just the technical muscles that manufacturers show their big models to the industry in the most intuitive way.
The big model supported by them is a common pursuit to enter the vertical field to realize commercialization after exerting influence.
In the big era of AIGC, companies that are ahead are already exploring more scenarios.
In the matter of commercialization, manufacturers have different tendencies-intuitively, they are divided into two paths: entering B-end enterprises to provide services and innovating products for C-end.
Whether to go to the B end or the C end has also triggered a series of discussions and even debates.
The source of this wave, OpenAI, tends to provide services to enterprises and charge subscription fees to quickly achieve hundreds of millions of dollars in revenue targets.
This has also become the choice of many domestic manufacturers.
In fact, the original meaning of the big model is "pre-trained large parameter algorithm model", and its essential meaning is embodied in "pre-training", which can quickly adapt to the corresponding scene after fine-tuning, and can form a landing application without providing a lot of data feeding, thus improving the ability of AI and accelerating the speed of AI innovation.
However, from the technical landing of large models, verticalization is an innate gene.
In Europe and America, where enterprise-level services are very strong, OpenAI gives priority to collecting subscription fees for enterprises.
In the China market, although with the rapid development of To B digital and intelligent services, manufacturers eventually need to provide services to enterprises in an all-round way. They also have another choice, that is, relying on the booming C-end market to demonstrate AIGC’s capabilities in various C-end products or services.
Intelligent terminals, which have been developed for many years in the past, have three advantages of commercialization: large market capacity, value potential of in-depth services, and easy grafting of AIGC capabilities through voice assistants.
Simply put, the market is big, the potential is deep, it is easy to get started, and it is easy to get universal choice.
This point has also been proved by the practice of research institutions and manufacturers.
In "AIGC and ChatGPT: Present Situation, Trend and Governance", China ICT Institute believes that intelligent voice assistant will be an important application direction of AIGC;
Guosheng Securities mentioned in its research report that the intelligent assistant will become a super application in the second half of the year. The large model allows human beings to command the intelligent assistant installed in various devices such as smart speakers, headphones and cars to help them complete various tasks.
The bottom support of companion products such as smart glasses is intelligent voice assistant.
Previously, Baidu’s smart speakers had been reported to be equipped with AIGC capability. Once released, iFLYTEK Spark model was directly applied to its learning machine hardware products, so that students and users could directly chat with the learning machine in English.
Before the introduction of smart glasses, Tmall Elf also unveiled the first intelligent hardware demo connected to the thousand questions model in April.
All intelligent hardware terminals that have appeared in the market in the past can be "recreated".
Speakers, headphones, cars, mobile phones, and even all smart homes can be "AIGC-oriented" by carrying voice assistants.
It is no exaggeration for Guosheng Securities to say that the assistant /AI partner will be the next super application.
However, further, in many smart terminals, not all products can be quickly highlighted.
Those portable terminals that inherit the concept of "wearable devices" will take the lead in showing new value through AIGC.
2
Three conditions are complete,
Large-scale model reengineering intelligent terminal
The landing of an AIGC intelligent terminal depends on three aspects:
Whether the product attributes fit AIGC and can collide with the maximum value directly determines the market acceptance;
Whether technology can keep up, which determines the functional quality of products;
Whether the track has a strong motivation to integrate AIGC determines the long-term commercialization depth.
The portable terminal represented by smart glasses is undoubtedly more suitable for these requirements among many intelligent terminals.
1. Product attributes: Sound brings more direct value.
In the Marvel Comics movie Iron Man, Jarvis, the AI housekeeper, helps Iron Man to manage all kinds of affairs.
This setting has been used by many movies. Almost every science fiction film can see an inseparable device accompanying the protagonist, providing all kinds of help, and even simply "accepting the talk" without performing the task.
Andrej Karpathy, the founding member of OpenAI, also said that he wanted to create a Jarvis.
Voice interaction+omnipotence+portability has become the common feature of Jarvis.
In all intelligent terminal devices, due to various restrictions, voice interaction has become the most commonly used, convenient and direct way to generate AIGC value.
And "carry-on" makes this value realization more frequent, and makes the value of "assistant" more closely fit with daily life.
After all, people’s range of activities is always changing, but their portable devices are always around.
It can be seen that although Tmall Elf’s smart glasses are more like AR glasses in product form, at least for the moment, it is essentially a sound interaction+portable way, and it is an earphone hidden by a "sci-fi" way.
In its report, Guosheng Securities also believes that the most suitable voice intelligent carrier is headphones in outdoor, traveling or scenes that require privacy. "Because of the limitations of human physical conditions, headphones are high-quality entrances in the future visible dimension."
Go to work to "fish", hold meetings to help record meeting minutes, go shopping to help find good places to eat, drink and have fun, and even talk late at night and become good friends of social fear users …
What Tmall Elf is trying to create is the inseparable life and work partner in science fiction movies.
2. Technology supply: The whole chain technology has been reserved.
At the conference site of each large model product, the normal demonstration process is often to input text and then AIGC creates an answer to present it.
This brings a problem: when AIGC enters the portable terminal and needs voice interaction, in addition to the capabilities of the big model itself, it also needs a series of technical support to form a technical chain:
In the perceptual stage, ASR technology converts the heard speech into text and inputs it to the large model;
Decision+creation stage, NLP (large model ontology), which creates content output after receiving information;
In the expression stage, TTS technology transforms the text into natural and smooth speech and delivers it to users.
For smart glasses, this is not over. To bring a private sound field, we need to cooperate with sophisticated bone conduction technology.
Only players with a series of technical reserves are eligible for the table. There is no chance of a dark horse here.
On the one hand, Tmall Elf’s smart glasses need to be equipped with the scene application of thousands of models, on the other hand, they can’t be separated from the long-term accumulation of smart speakers in ASR, NLP and TTS technologies over the years.
As an innovative product, the research and development of new engineering technology behind it is also worthy of attention.
For example, the bone conduction technology used this time adopts the so-called "patented vibrator technology", which optimizes the connection mode of the bone conduction sound device and makes the vibration response more stable, less noise and less distortion.
Moreover, the implanted components need to be miniaturized to ensure that they can be placed in 9.9mm temples.
Both technology and experience refuse to be bulky.
Of course, with this series of technical reserves, Tmall Elf is not only used in glasses, but also in wearable devices such as headphones and helmets in the future.
If smart glasses are successful, we can expect more AIGC innovations of portable terminals in the future.
3. Change demand: AIGC meets the needs of the track to break through itself.
For a long time, wearable devices have been in a situation where the concept is hot but the product progress is slow.
After the portable terminal is equipped with a large model and the AIGC capability is realized, the possibility of more value creation appears.
Accompanying, chatting, recommending restaurants, online shopping, office work, information retrieval … all can be provided with a portable device.
Liu Shuo, deputy director of the Content Technology Department of the Institute of Cloud Computing and Big Data of China ICT Institute, called the value change of AIGC to the original track "destructive innovation"-the value of AIGC to the portable terminal is no longer based on the optimization of the original product logic, but a kind of "reconstruction" after breaking.
Such is the change brought by smart glasses to AR, and in turn, it is also driven by this change demand.
Taking the tourism scene as an example, it is remarkable that AR glasses can provide corresponding information for decision-making after seeing a hotel. Now, the built-in "little friends" can analyze users’ needs, retrieve network information, make suggestions for users, and even help users to place orders and complete the construction of tourism plans through voice.
There is reason to believe that more tepid portable terminals and even smart terminal tracks are in urgent need of such innovation.
three
AIGC competition of portable terminal,
Still returning to the industry itself.
The portable terminal equipped with AIGC has brought brand-new efficiency and scene demand satisfaction, stimulated new market space, not only provided specific services for users, but also created an ecosystem with multi-dimensional participation.
Intelligent terminals have been providing users with rich services in various ways and paths in various work and life scenarios.
The addition of AIGC may also make the competition of smart terminal products in the past get rid of homogenization.
For example, in the product competition of smart speakers, a typical intelligent terminal of traditional voice interaction, it is hard to say that there are too many differences among manufacturers-responding to users’ needs in the form of databases and knowledge maps, and each product is basically "you have me".
However, when the big model becomes the new base infrastructure and the AIGC capability is delivered to the portable terminal, the capability of each product will be greatly differentiated. Product competition will become hard power competition.
* The pictures in this article are all from the Internet.
This article first appeared on WeChat WeChat official account: Intelligent Relativity. The content of the article belongs to the author’s personal opinion and does not represent Hexun.com’s position. Investors should operate accordingly, at their own risk.
关于作者