With the blessing of funds, will artificial intelligence still be the main line of the market in 2024?

"Weekly Capital Market Observation" is a live broadcast column of the interpretation of capital market news launched by Titanium Media, every Monday from 12:30 noon to 13:30, in an hour, to help everyone catch the major events of the capital market in one week!

In this episode, Yu Hao, the director of "Weekly Capital Market Observation", invited Chen Chen, a partner of Analysys Analytics Research, to review the overall development of the artificial intelligence industry in 2023 and share his views on new opportunities in the artificial intelligence industry in 2024.

Weekly Capital Market Performance Review (12.18-12.22)

Last week, the three major A-share stock indexes all fell, with the Shanghai Composite Index down 0.94%, the Shenzhen Component Index down 1.75%, and the ChiNext Index down 1.23%. Last week, northbound funds sold a net of 2.236 billion yuan, and southbound funds bought a net of 145 million yuan. In terms of industry preference, the top three industries in which northbound funds increased their holdings were batteries, photovoltaic equipment and automobiles; the top three industries in which southbound funds increased their holdings were the energy industry, consumer service industry and insurance industry; and the top five industries in which northbound funds increased their holdings in individual stocks last week were Shenzhen Textile A, Cixing Shares, Duolun Technology, Yunyi Electric and Hengxin Oriental. The top 5 companies in terms of market value increase in southbound funds last week were CNOOC, China Property Insurance, Kingdee International, SenseTime and China Shenhua.

Both major indices of Hong Kong stocks fell last week, with the Hang Seng Index down 2.69% and the Hang Seng Tech Index down 6.15%.

All three major U.S. stock indexes posted gains last week, with the Dow up 0.22% for the week, the S&P 500 up 0.75%, and the Nasdaq up 1.21%.

Last week, a total of 4 new stocks were listed on the Shanghai and Shenzhen stock exchanges, of which Hongsheng Huayuan rose the highest on the day of listing, 340.59%, and Hong Kong stocks had 7 new stocks listed last week, of which Fanyuan International rose the highest on the day of listing, 11.11%.

In December 2023, Analysys' annual report "Chinese Artificial Intelligence Industry Application Atlas 2023" (hereinafter referred to as "Atlas") was officially released, and the "Atlas" mentioned that the development of artificial intelligence has entered the era of generative artificial intelligence, what is the era of generative artificial intelligence? What are the characteristics of this stage?

Chen Chen, a research partner at Analysys, said that the ultimate goal of Atlas is to return to the essence of artificial intelligence and see the true value it brings to all walks of life. Therefore, at the beginning of "Atlas", the current stage of artificial intelligence development is divided.

The era of generative AI has the following important characteristics:

1. There has been a significant evolution in the way humans interact with each other, redefining the way humans and technology interact. It is not only a change in technology, but also a comprehensive upgrade in experience.

2. One of the most significant capabilities of generative AI is that it can solve problems that it has never seen or rarely encountered by learning Xi and summarizing the original data.

Chen Chen pointed out that many people are now discussing whether human beings have begun to move towards the era of general artificial intelligence, but in the "Atlas", Analysys still defines the current era as the AGI 0.1 stage, that is, the stage of interactive revolution. When entering the stage of AGI 1.0, AI will not only have world knowledge, but also everyone's personal knowledge, and it will also usher in a knowledge revolution.

At present, what are the bottlenecks encountered in the development of large AI models?

Chen Chen pointed out that the current problems faced by the development of AI large models are mainly in the following aspects:

1. The cost problem is also the core problem at present. In addition to the high training costs faced by manufacturers who produce large models, those industry customers who use open source large models + industry-specific data for incremental training also need to pay high fees. In addition, there is a very large gap between the perception of large model manufacturers and industry customers on the application side about the cost of deploying a large model. In the future, it is still necessary to find some more effective ways to reduce the cost of training and inference, so that it is possible to better help and accelerate the implementation of large models.

2. At present, including ChatGPT, even the latest version, the data used is not real-time. At present, the mainstream large model training method is still based on the use of static data for learning Xi training. However, some professional fields need to use time-sensitive data and some proprietary data to generate auxiliary decisions, so this point needs to be broken when entering these professional fields.

3. Long-term memory problems. At present, this problem can be solved by some technical means, such as increasing the context capacity and using vectorization. In the future, more ways to solve the limitations from the technical level need to be explored.

4. When customers in professional fields such as finance and medical care use large models, they will have very high requirements for the interpretability and data traceability of the models, which is also the dilemma faced by customers in these industries in the process of large models going deep into vertical fields.

5. Data privacy issues, how to align AI models with universal values and industry standards, etc.

According to the Atlas, generative AI accelerated by large models has penetrated into multiple scenarios. At present, which industries are benefiting from the application of AI technology, and what are the representative companies in these industries?

Chen Chen said that the "Atlas" mainly selects six industries: industrial manufacturing, retail, finance, medical and health, entertainment and digital government, because these industries have relatively obvious industry attributes, and we can deeply explore whether the application of AI can really bring significant changes to these industries.

When talking about the selection of representative cases in various industries, Chen Chen introduced that one of the perspectives of selecting cases is whether there is a good benefit ratio, of course, the cases are inexhaustible, and the "Atlas" only selects a small part of them, and there are many very good cases that I hope to discuss with more industry partners.

1. Manufacturing industry: Haier Kaos's industrial Internet platform helps build an intelligent factory hub through industrial models to drive intelligent factory upgrades, and Huawei's industrial AI quality inspection technology has also been applied to many manufacturing production lines.

2. Medical industry: Focusing on the medical and health industry, Tencent has fully accumulated medical imaging and medical information, providing overall solutions for different types of medical institutions to upgrade their digital intelligence

3. Entertainment industry: CMGE uses AI applications to improve the efficiency of game research and development and develop 3D assets.

4. Digital government: 360 Group's tax industry model helps smart taxation, and SenseTime's AI device empowers urban fine governance.

What are the different challenges that different industries face when introducing AI technology?

Chen Chen pointed out that the AMC curve is constructed in the Atlas based on the development characteristics of the industry, and the maturity of AI applications in different industries can be understood:

1. Agriculture, energy and other industries are still in the exploratory stage. The digital infrastructure of this type of industry is still in the process of being built. What needs to be strengthened is how to better precipitate data resources to provide a foundation for the future upgrade of digital intelligence.

2. Manufacturing, finance and other industries now have basic digital capabilities, and the problems that need to be solved include how to build a more credible and compliant model based on industry characteristics and actual business scenario requirements, and how to roll out applications in a large area of the entire business process.

3. In industries such as entertainment and e-commerce, AI applications in these industries are developing rapidly, and more important considerations in the future may be how to further enrich content and provide more personalized experiences in order to better achieve large-scale industrial-level applications.

Recently, Apple's first-generation MR (mixed reality) product Vision Pro ushered in new progress, which has attracted widespread attention in the market, as one of the application ends of artificial intelligence technology, how do you see the development prospects of the MR industry?

Chen Chen said that now when everyone talks about MR, they will also mention AR and VR at the same time. At present, VR and AR are actually a bit of a "fire and ice" situation, VR is a little cold, but the current sales of consumer-grade AR glasses seem to be good.

VR emphasizes more on the immersive experience in the virtual world, but its problem lies in the lack of core application scenarios and excellent content, the use experience has not met expectations, and the corresponding ecology has not been done.

When it comes to AR, Chen Chen believes that AR emphasizes augmented reality and amplifying the senses, and there are already core scenarios such as travel and business trips, so it is still relatively popular on the C-side.

MR emphasizes the interaction between virtual and real, which is also the biggest feature that distinguishes it from AR and VR. Some bloggers who have tried Apple's Vision Pro think that this product integrates virtual objects with the real environment very well, and this may actually be a good point where MR can be combined with AI, in some to B scenarios, there can be a lot worth exploring, such as remote industrial inspection, telemedicine assistance, cultural tourism scene and interaction in teaching, etc.

In 2024, what new developments and opportunities do you think the AI industry will look forward to, and what segments are worth paying attention to?

Chen Chen believes that the industry application and corresponding maturity of artificial intelligence will continue to attract attention next year, and there will be a lot of room for exploration in the future. At present, the industrialization level of the content industry may still be more inclined to the application of some point links, and in the future, it may be more powerful in the application of the whole process, so as to improve the industrialization level of the content industry.

In the future, some small and medium-sized enterprises and individual creators may have the opportunity to break through, and with the blessing of AI technology, they will have more space to conceive some more creative and better experience content.

When it comes to the subdivisions worth paying attention to, Chen Chen suggested that you can pay attention to the following aspects:

1. Video generation may have a breakthrough in 2024;

2. Existing applications and devices, including apps, office applications, smart homes, in-vehicle devices, etc., will be upgraded with AI technology in more products.

3. Based on the native capabilities of generative AI, more related applications will be born.

For more capital market content and views sharing, please search for the "Weekly Capital Market Observation" column on the Titanium Media App and watch the live broadcast of the program.

Disclaimer: The news of this live broadcast column comes from public information, all content and suggestions are for reference only, and should not be used as the only reference factor for investment decisions, investors should make investment decisions independently and bear their own investment risks. Investment is risky, and you need to be cautious when entering the market.

(This article was first published in Titanium Media APP, author|Content Operation Editor Yu Hao)