Source: NewPosition

Image Source: Generated by Wujie AI

On September 15th, it was reported by the media that Alibaba's "Tmall Genie" will be renamed, and new products with the XGENIE brand have already appeared in some channels. The response from Alibaba's relevant person in charge was, "Indeed, there are plans to launch a completely new brand, strengthening the genie as a personalized feature for users."

Behind the action of "changing the name and retraining" is a new wave brought to the dormant smart speaker industry by the emergence of generative large models.

Last year, the domestic smart speaker shipments were only 26 million units, and the market structure showed a high degree of concentration, with the combined market share of the top three players, Xiaodu, Xiaoai, and Tmall Genie, exceeding 90%. However, the overall market share of these top players is not significantly different, with Baidu at 35%, Xiaomi at 31%, and Tmall Genie at 27%. The result of the three-way split means that none of them have crossed the threshold of 10 million units shipped.

This performance, placed in the smart speaker race that was once seen as the next generation key entry point for the Internet, even if it cannot be said that the bubble has completely burst, it has indeed receded. Looking back today, Alibaba's strategic layout around Tmall Genie with an investment of 10 billion yuan in 2020 probably does not match the cost-effectiveness of what Li Yanhong did with 20 billion yuan for Nuomi back then.

However, with the emergence of larger models, the smart speaker industry seems to have the possibility of a resurgence.

First, on February 9th, Xiaodu announced the integration of Wenxin Yiyu to create the AI model "Xiaodu Lingji" for smart device scenarios, and will apply it to all Xiaodu products.

Then, at the Alibaba Cloud Summit on April 11th, Tmall Genie also announced that it will combine Tongyi Qianwen. Zhang Yong, who also serves as the CEO of Alibaba Cloud, stated that in the future, Tmall Genie hopes to become a "personalized intelligent assistant" and introduced the "Niaoniao" voice and personality that imitates the stand-up comedian Niaoniao.

Although Xiaomi's progress in large models is slightly slower than that of the Internet giants, it has also started testing the AI large model version of Xiaoai Assistant last month.

However, although large models are hot, the future of smart speakers does not seem so bright.

Smart speakers are the entry point, and smart home management comes first

When Echo was launched in 2014, it did not sell well, at least not in terms of sales. Initially, Amazon only allowed Prime members or invited users to use the product. However, shortly after the launch of Echo, major companies including Google and Apple decided to create similar products because they believed that smart speakers would become the center of users' home life, especially as an interface for various smart home devices.

This logic was later recognized by many domestic companies, and they successively launched their own smart speaker products, causing a moment of confusion for consumers. Later, the situation followed the same path as other segments of the Internet technology industry. Baidu, Xiaomi, and Alibaba relied on strong financial resources to complete the encirclement of other players and maintain the "tri-polar" market structure to this day.

However, the regret is that the outstanding players did not reap the fruits of victory. According to data from Luotu Technology, the total channel sales of smart speakers in China in 2022 was only 26.31 million units, a year-on-year decrease of 28%. In the first half of this year, the industry's sluggish trend was maintained, with total channel sales of 11.48 million units, a year-on-year decrease of 19%; the corresponding sales revenue was 3.37 billion yuan, a year-on-year decrease of 20%.

One possible explanation for the decline in the smart speaker market is that the target audience it hopes to serve is not as large as expected, and consumers are not as enthusiastic about smart home products.

From the perspective of the development history of the domestic Internet, the pace of consumer Internet is much faster than that of industrial Internet.

All the national-level software that users can see on their mobile phones have experienced an extremely steep growth curve in the process of popularization among C-end users. WeChat and Douyin are typical examples. For instance, WeChat was launched in 2011, and by the end of 2016, its monthly active users had reached nearly 900 million, basically equivalent to covering the entire population of Chinese mobile Internet users at that time.

In contrast, the pace of industrial Internet is much slower, even in the most dynamic cloud computing industry. After the initial batch of Internet companies achieved digitalization and cloud migration, the subsequent expansion of traditional industries could only be pushed forward by the large sales teams of major cloud vendors, to the extent that all cloud vendors had to focus on the enterprise end, as it is currently the most stable source of business.

However, in a specific segment, the positions of consumer Internet and industrial Internet are exactly the opposite, that is, smart devices or the Internet of Things (IoT).

Companies with factories generally have strong motivation to install various sensors and robots in the workshop, both to reduce labor costs and to significantly improve production efficiency and safety. For example, a key support point for Tesla's success is the use of a large number of automated equipment and software systems in its super factories, and Musk has even introduced the concept of autonomous driving to the assembly line. Today, the penetration rate of smart devices in the domestic automobile manufacturing industry has reached 36.5%.

In contrast, C-end smart devices, which we usually refer to as smart home devices, are significantly lagging behind in terms of promotion and application. In a research report from April by Founder Securities, it cited data from Statista, giving the current penetration rate of smart homes in China as only 13%. Considering that many so-called smart home devices are in a state of disuse, the actual level of utilization may be even lower than this data.

The fundamental reason for this differentiation is that smart devices in factories are a "necessity," while smart devices at home may not even be considered "useful."

Taking the smart speaker focused on in this article as an example, most consumers who bring this product home use it more as a "speaker" than as "smart," such as telling stories and listening to music. The results of a survey by Rakuten Insight show that as many as 79% of Chinese respondents who own smart speakers stated that they mainly use smart speakers to search for music and play songs.

Due to the aggressive competition strategies adopted by various companies in the past, the prices of smart speakers are generally difficult to cover the costs at all levels. After the dream of "reaching for the stars and the sea" was shattered, some unconventional operations had to be implemented to recover. For example, a few years ago, Tmall Genie allowed free listening to some member songs from certain apps, but now, even if you have subscribed to a music app, you need to further subscribe to Tmall Genie's membership to listen to songs.

Other smart home devices are often not much different from smart speakers, and they are not smart enough to create essential value among consumer groups. Considering that household products are generally durable, if there is no necessity, consumers will take a relatively long time to adopt new products. IDC estimates that the global shipment volume of smart home devices will decrease by 2.6% in 2022. (Note that this decline should not be compared with that of smartphones. Smartphones have already reached saturation, but a large number of traditional home devices have not been replaced.)

Another issue to consider is the compatibility of smart speakers with various smart home devices from different brands. Alibaba has allocated 10 billion yuan, with a focus on building an ecosystem around Tmall Genie. However, since home appliance manufacturers tend to avoid binding with a specific technology company, many products support access to smart speakers from different brands, and basic compatibility is generally not a problem.

Large models are revolutionary, revolutionizing the fate of smart speakers

Of course, the recent fate of smart speakers cannot simply be attributed to the overall lackluster situation of the smart home market. The main reason is still that smart speakers themselves are not smart enough. Many times, like other artificial intelligence products, smart speakers exhibit behavior that leans more towards "artificially disabled."

Usually, users can only give it very simple commands, such as playing music, checking the weather, controlling lights, or telling bedtime stories to children. Once the structure of the voice command is slightly complex, or involves multi-round interactive communication, the performance of smart speakers becomes a disaster.

The work of smart speakers involves three processes: converting the user's voice into text commands, understanding the meaning of the commands and processing them, and then converting the processed results into text output. The above three processes correspond to three technologies: automatic speech recognition (ASR), natural language processing (NLP), and text-to-speech (TTS).

Among them, the conversion between the beginning and end of voice and text is already very mature, and the root cause of the "intellectual disability" of smart speakers lies in the middle natural language processing. The emergence of generative large models has made a leap in the computer's natural language processing ability. In general scenarios, GPT has shown conversational abilities that have led some to start discussing the distance to general artificial intelligence (AGI).

So, there is no doubt that generative large models can indeed greatly compensate for the lack of "intelligence" in existing smart speakers. Therefore, as mentioned at the beginning of this article in "New Position," the existing top smart speaker manufacturers are taking action in this regard. And from the timeline, everyone is extremely excited and anxious. For example, when Baidu announced on February 9th that it would integrate Wenxin Yiyu into Xiaodu, Wenxin Yiyu was not even aware of it.

Large models may bring about a revolution for smart speakers, and this indeed seems to be a valid logic. In addition to the enthusiasm of the manufacturers of various smart speakers, the industry media focusing on this direction is also filled with similar views.

However, in the view of "New Position," although large models have solved the biggest problem faced by traditional smart speakers, they have also completely deconstructed the envisioned application scenarios of smart speakers.

Since its inception, the most prominent label attached to smart speakers has been "entry point," another thing with the same label is the smartphone. If large models are truly intelligent enough, then the key to determining the fate of smart speakers is whether they can compete with smartphones in this "entry point" role.

The answer is obvious. Except for the output power, all the tasks that smart speakers can perform can also be done by smartphones. Conversely, the tasks that smartphones can accomplish and their close connection with users are far beyond what smart speakers can achieve. In a scenario that the author is aware of, some users have multiple smart speakers in different areas of their homes, such as the living room and bedroom. If you need to give commands to smart speakers in an elegant way in various areas of your home, this is necessary.

So, smart speakers basically have no differentiation advantage compared to smartphones.

After the emergence of large models, due to the amazing performance of large models in general intelligence, people will obviously tend to regard large models as true personal assistants or companions, and then use them to solve as many problems as possible. In this case, the role of smart speakers will be accelerated to merge with the personal assistant of smartphones, because that is the truly irreplaceable entry point.

Ultimately, if everyone is "artificially disabled" and the user's demand is not high, then I will not replace you so quickly; but if there is real artificial intelligence and the user's demand arises, then they will use this artificial intelligence assistant to help with tasks such as writing PPTs, searching for information, communicating for emotional comfort, and of course, controlling household appliances such as refrigerators, air conditioners, and light bulbs, rather than being limited to a specific scenario.

Concluding Remarks

An interesting fact is that before launching Echo, Amazon actually had a higher-priority smartphone project called Fire Phone. If it weren't for the failure of the smartphone project, it would be questionable whether Amazon would have launched the Echo product later.

After almost a decade, the smart speaker industry has gone through a complete cycle of rise and fall. Before the outbreak of generative large models, the biggest reason this industry still existed was not user demand, because if users needed something, they would be willing to pay a fair price for it, rather than having manufacturers give it away.

As the smart speaker industry was on the verge of gradually fading away due to inertia, large models arrived. But it is not a stimulant, but a death knell. Smart speakers were redundant from the beginning, and with the arrival of large models, they are even more redundant.

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。