Some use other people’s technology. The “house” that was originally thought to be created according to the algorithm is ideal, but it is often not.
Especially in terms of speech, there are a lot of semantic understandings that are difficult to standardize. At least eight to ten dimensions must be calculated, not an algorithm.
In addition, the cost of R&D investment is relatively high, and a large amount of capital is required. The first is the need to study this industry. The preliminary research costs are relatively high. Secondly, the research products are good. Although there are many application scenarios, the real commercialization is a long process. This process requires a lot of money.
For the current industry challenges, Analysys Research Center analyst He Wenqian believes that although face recognition technology can meet certain commercial standards, it is still used in areas such as payment and other requirements for robustness and accuracy. There are certain technical bottlenecks. At the same time, the level of intelligence that natural language understanding technology can achieve is still not enough in the Turing test to give it a smooth experience.
She believes that the cause of this problem, on the one hand, is related to the lack of data, the current deep learning framework requires a large number of data feeding models to improve accuracy.
Taking 3D vision with high accuracy and robustness as an example, the current 3D face image data is insufficient, which affects the large-scale promotion of the technology; on the other hand, the cost of AI landing is a major factor. Both visual and voice involve hardware devices such as sensors, processors, servers, and the most current AI talent. Under relatively rough and primitive implementations, corresponding to relatively high costs, the risk and cost of selecting smart interactive products may hinder them from trying.
However, compared with technology and input costs, Xu Lixin believes that the market needs education guidance. Due to the improvement of service demand standards for B-end customers, the output of AI intelligent interactive services needs to conform to market changes.
At present, the main consumer power in some areas has gradually shifted from after 70, 80 to 90, and after 00, their perception of service has changed a lot, so the product design should be more in line with this batch of consumers. The needs, such as more able to show personality, more interactive, focus on consumer experience, etc.; and for enterprises, because the market demand has changed, the internal corresponding service groups will also change.
Industry sources said that the current transition of intelligent interaction methods in most scenarios often relies on AI companies to export their own AI technology to enterprises or entities to help them realize the intelligence of a certain interactive link. This requires AI companies to better understand the needs and changes of the B-end market, and even better understand the changes of C-side users than the B-end, and guide them. In the technology research and development and strategic adjustment to keep up with the trend of the industry, otherwise it will soon be eliminated.
5G boosts B-side intelligent interactive application to speed up capital favoring unicorn companies
Although the development of intelligent interaction faces many challengesHowever, the arrival of the 5G era has accelerated the application of intelligent interaction on the B side.
Song Jin believes that the biggest development opportunities for voice interaction are in consumer IOT devices, as well as on the B side. A relatively positive trend now is that 5G will promote the development of IOT equipment. Smart speakers, service robots, smart cars, smart glasses and smart headphones are the most promising applications for voice interaction.
Smart speakers are the best example of voice interaction. In the first half of 2019, China’s growth in China exceeded 200%; and automakers are also striving to create full-vehicle voice interaction; smart glasses, smart headsets, and voice separation. Empty operations are also more convenient than touch.
Xu Lixin believes that the application of intelligent interaction in the future can be reflected in the services of all walks of life, and intelligent interactive services will become a major trend. In her view, each industry can have a space for in-depth research and application in the improvement of services.
Take the range hood as an example, you can use it as an intelligent robot. After entering the kitchen, you can tell the recipe you want to make. The range hood will tell you how to make this dish. When you open the refrigerator, you will find this. The ingredients of the dish are not enough. If the order is given to the range hood, the order will be automatically placed. After 10 minutes, fresh ingredients will be delivered to the home.
The future development trend is that products in certain service segments will be smarter and more consumer-friendly in addition to daily services. But the problem is not small, and more partners need to participate in such a cooperation platform to .
Still taking a range hood as an example, at least one chip needs to be implanted so that the robot can read the menu, and how the menu interacts with the AI technology provider’s back-end customer service center, and how to master the family members used. Big data, preferences, etc., all require professional support for each link of the entire chain.
Therefore, AI companies must establish an ecosystem when exporting technology. By building their own smart aisle, all data can be collected on this platform, not only for AI companies, but also for partners of the entire ecosystem. In order to enable smart interactions to be applied to multiple scenarios, the business model will really land.
In Xu Lixin’s view, the capital market is more optimistic about the AI technology service companies that have achieved a certain scale on the B side, because the profit model is clearer than the Internet companies, and with the rapid development of this market, capital Has begun to increase the layout strength. For start-up companies in this industry, capital is more rational, and early project financing is more difficult. This requires new companies to have clear profit models and technical strengths. Therefore, the bottom of service technology is built for such startups. more important.
Whale-Kwon-dong sees data shows that among the 2,17 human intelligence projects it included, 658 projects in the first half of 2018-2019 were successfully financed, totaling approximately 88.6 billion yuan. From the perspective of financing, 2In the first half of the year, 018-2019, the number of financing events in the B round and afterwards accounted for about 24% of the total number of artificial intelligence financing events, which was significantly higher than the 15% in 2017.
From the perspective of distribution, large-scale financing is concentrated in core technology vendors such as computer vision, robotics, chips, and autopilot, and the head unicorn financing in each segment is not reduced.
For the future development of intelligent interaction, He Wenqian said that the first is to be optimistic about the development of AI interaction mode. The improvement of upstream and downstream supply chain will enrich supply and reduce costs. Secondly, AI open platform, especially automated deep learning platform will Greatly reduce the technical threshold of AI application, and then reduce the labor cost; finally, in the gradual landing process of AI, the increase of data volume will continue to optimize the effect of AI landing, intelligent interaction will become more and more original biochemical, gradually replace the current Some interactions subvert the existing industry form.