Users first, then customers, to make AI truly a positive business cycle

Scenario implementation has always been the number one issue in the AI ​​chip industry.

AI chip companies have emerged in large numbers since 2015. After several years of market competition, the camps have gradually differentiated: on the one hand, there is a constant stream of news of "Mercury retrograde", with department layoffs, business adjustments, wafer failures, and difficulties in going public; on the other hand, some AI chip companies have received multiple rounds of financing, their products have continued to iterate and upgrade, and they have seen new growth on the market.

What exactly caused the simultaneous "advance" and "retreat" of AI chips? Interpreting the "watershed" behind this may point directly to the core and logic of how to implement AI.

Difficulty in implementation, difficulty in user experience

Lu Yong, CEO of Tanjing Technology, believes that in theory, AI is "all-powerful" in voice and image recognition, but as a startup company, it is important to calm down and think about whether this solution is a timely help or just a icing on the cake for users.

Take the smart home field as an example. This concept has been highly anticipated since its inception. However, more than ten years have passed, and the market share is still extremely low, and the product stickiness is far less than originally expected.

What exactly is the problem?

From the user's perspective, this problem is easy to understand. Since AI voice capabilities require the assistance of cloud computing power at the beginning of development, many people in the industry naturally think that cloud computing is the only direction for intelligence. However, the operation of smart products connected to the cloud is cumbersome. To meet the rigid requirement of Wi-Fi networking, multiple operations such as downloading APP, registering, setting, and connecting to a third-party central control are required.

"From the user's perspective, this is not user-friendly enough. If I want to turn on a light or air conditioner, I have to consider the network status first and then make various settings. This scares away many users, and they dare not bring such 'high-end' smart products home." Lu Yong believes that no matter how cool the functions of a smart home are, it is essentially still a household appliance.

Among the consumers of smart home appliances, only a very small number of users are interested in advanced technology, are proficient in configuring the network, and are willing to spend a lot of time studying the products. This type of users can be called geek users. But in fact, this type of users only accounts for a small part of the entire consumer group, and the vast majority of ordinary users do not have such IT capabilities.

When a new technology emerges, the most feared thing is falling into "cognitive bias". People who design products will substitute their own technical capabilities, and take it for granted that everyone should understand it. Such engineering thinking usually results in a large and comprehensive product, but the threshold for use is extremely high. This seems to be a curse of knowledge, making high technology a natural shield that abandons the most frequent users.

Users are afraid to buy and they don’t know how to use it. These two major problems have seriously affected the consumer experience. If we want to open the door to AI applications, we must break the ice of customers’ habitual thinking.

The end-side doctrine of exploration

In Lu Yong’s view, only when home appliances themselves have AI capabilities, do not rely on third-party equipment, and can achieve human-computer interaction without complicated settings, can they be truly “people’s intelligence”.

Looking back at the historical evolution of technology, we can find that computing power was first released from large computer rooms to personal PCs, and then from computers to mobile phones. All computing power has evolved from being centralized at first to each doing its own job at last. In the AI ​​era, it is against market value to concentrate all computing power in the cloud.

The advantage of end-side voice is that consumers do not need to configure the network, central control or gateway equipment, and can use it by plugging in. A single machine can achieve intelligence, which greatly reduces the learning threshold and usage threshold for users. In addition, as users pay more and more attention to privacy, personal data privacy has also become a consideration for users when purchasing and using smart homes.

Henry Ford once said: "If I had asked consumers what they wanted, they would have told me 'a faster horse!'". But when users drive a car, they will not care about the fastest horse in the world. - When users truly experience the feeling of offline voice control of YYDS, manual operation will no longer be "welcomed".

This is equivalent to the cultivation of user habits, just like the last mile mentioned in the express delivery industry, and for smart home voice applications, it is the last centimeter of distance.

Cultivation has always been easier said than done, and Tanjing’s breakthrough does not rely on simple preaching, but on a solid package of high-dimensional systematic services.

"Tanjing VUI must first achieve success for users and then for customers," Lu Yong said with deep feeling.

In the past, most chip manufacturers only provided chips that met customers' PPA indicators. The next level was to provide Total Solution, and the next level was to open SDK interfaces. However, regarding AI voice chips, Tanjing not only provides a complete set of solutions including chips and algorithms, as well as SDK interfaces, but also further "upgrades" its thinking, considering how to improve user experience, and superimposing it on the top level to consider how to achieve customer value.

This is because many B-side customers have not thought clearly about the direction and scope of AI applications, and are just following the trend. Some companies tend to adopt a customized route, and just copy whatever needs are raised by home appliance customers or solution providers. Tanjing has chosen another dimension of "tactical forward movement", starting from the user side instead of the industry side, and refining and improving the solution from the user's perspective.

In terms of the implementation of more specific home appliance products, Tanjing chose to start with environmental control appliances such as lamps, air conditioners, fans, and other home appliance solutions, and entered into a full range of home appliance solutions. Lu Yong explained that for environmental products, users have the strongest personal experience of intelligence. The changes in light brightness or the changes in the temperature of air are derived from people's basic survival needs. They are the most in-demand products, and are used very frequently, and it is easy to develop a habit of using them.

"After paving the way in the past few years, some industry leaders are now actively pushing forward the implementation of offline voice, because the value we create is passed on to customers, and customers also recognize the deep logic of the value. Under this changing situation, the smart home industry in 2022 will accept offline voice as an important functional point, and will usher in a major change." Lu Yong concluded.

New intelligent core on the end opens up a new blue ocean

Smart home may be a large market second only to smartphones. Lu Yong analyzed that there are about 500 million to 1 billion home appliances, about 150 million air conditioners, 80 million televisions, 300 million lighting fixtures, and about 300 to 500 million small home appliances. This market base is huge.

Once the door to the client-side voice market is opened, it will usher in a massive new blue ocean market.

Lu Yong explained this judgment based on changes in several aspects: First, more and more users are beginning to get used to using voice interaction instead of remote controls, and the market demand for low-power devices with end-side voice interaction functions is growing rapidly.

The second is the field of smart home, from lamps, refrigerators, washing machines, and air conditioners, to small appliances, white appliances and other usage scenarios, these fields are all progressing in parallel.

Third, initially only large companies tried voice AI functions, and they used them in some high-end products to explore the market. But since last year, the pace of voice intelligence in low- and mid-end products has accelerated, and even some small and medium-sized brands and even white-label products have taken the initiative to seek differentiation in the voice AI experience in order to compete and differentiate with large brands.

Lu Yong believes that the penetration of intelligent voice will start with devices with remote controls, followed by terminal devices with touch and buttons.

This is where the fission effect occurs.

"After a customer launches a product, he will find that the market feedback is very good, so he will launch more categories, resulting in internal fission; then there will be external fission, when a large customer or an industry dark horse adopts edge-to-edge voice, other customers in the industry will catch up." After four years of hard work, Tanjing Technology has ushered in a new harvest period.

According to Lu Yong, in 2021, Tanjing released the second generation of voice chips, including VOI311 (light version), VOI621 (upgraded version) and VOI721 (enhanced version). As soon as they were launched, many customers placed orders one after another, and several major customers took the initiative to find Tanjing to sign a strategic cooperation agreement, hoping to ensure supply.

The road ahead

After being tested by the market, Tanjing’s next step will be more stable and long-term.

Lu Yong mentioned that Tanjing will continue to optimize in the fields of products and algorithms. First, product iteration at the chip level, and second, continuous in-depth development at the algorithm level to launch a more practical and cost-effective voice solution.