From general large model to vertical large model, from single mode to multimodal integration, AI generation technology continues to iterate, accelerating application landing and business model innovation, which is the general trend.
Financial fraud is ST
ST Xiachuang Opening limit
In the morning trading on May 14, ST Xiachuang encountered a 20 cm down limit, and the number of down limit orders once exceeded 690000. This stock has been subject to other risk warnings since today.
ST Xiachuang's announcement shows that there are false records in the company's 2018 annual report and 2019 semi annual report. The Company falsely increased its operating revenue by 75 million yuan in its annual report in 2018, and by 47 million yuan in its semi annual report in 2019, totaling 122 million yuan.
In addition to ST Xiachuang, there were also early limit falls Medical aid for the people 、 Nandu Property 、 Nanhua Biology 、 Huarong Holding Etc.
Open AI Releases "The Best Ever"
Latest multimodal model GPT-4o
On May 13 local time, Open AI released a new flagship generative AI model called GPT-4o, and plans to "iterate" it into the company's products in the next few weeks.
It is reported that GPT-4o's text, reasoning and coding capabilities have reached the level of GPT-4 Turbo, and its speed is twice that of the previous generation of AI large model GPT-4 Turbo, but its cost is only half of that of GPT-4 Turbo. Video and audio functions have been improved. In the live demonstration, it seemed that a person was sitting next to it, and the conversation with human beings was natural and harmonious. It was completely inaudible that it was an AI. In just 232 milliseconds, GPT-4o could respond to the audio input. Moreover, it can take any combination of text, audio and image as input and output. This means that Open AI has taken an important step towards more natural human-computer interaction.
Altman, the founder of Open AI, said that GPT-4o is the best model of Open AI ever. It is smart, fast, and natural multimodal.
Guosheng Securities Research Newspaper believes that GPT-4o is a major progress towards more natural human-computer interaction. The new functions have brought new multimodal interaction capabilities. Through the new end-to-end model, it has achieved a new breakthrough in experience, which is expected to maximize the user experience in various terminals and benefit intelligent terminal agents robot Equal direction.
AI multimodal and long text application acceleration
Although multimodal AI is still in its infancy, many models have begun to emerge. Google's Gemini Ultra model exceeded GPT-4 in the multidisciplinary multimodal understanding and reasoning (MMMU) benchmark test, showing the potential of multimodal models. In order to remain competitive, more large-scale language model developers will follow up and develop multimodal functions. In addition, multimodal AI is expected to unlock new business opportunities, such as Artera's application in the healthcare field, Google's integration of Gemini into search, Ghost Autonomy's exploration in the field of automatic driving, and Meta's application of it to consumer devices such as smart glasses.
Since this year, the global AI big model end has been continuously upgraded, including overseas Sora, Llama3, domestic Kimi, Kunlun Tiangong AI, Step Star, etc. Huatai Securities It is pointed out that with the upgrading of model capabilities, 2C applications are expected to usher in accelerated development. The core of 2C application is to solve the problem of product performance and users' willingness to pay. With the optimization of the ability of the underlying basic model, the use effect of 2C application is significantly improved, and the application mode is accelerating.
Multimode is regarded as one of the important trends of AIGC industry in 2024. The China AIGC Application Panorama Report released by the qubit shows that the scale of China AIGC (generative artificial intelligence) application market will reach 20 billion yuan in 2024, trillion yuan in 2030, and the average annual compound growth rate from 2024 to 2028 will exceed 30%.
Multimodal AI Concept Unit
R&D intensity is much higher than that of A-share
In AI, a popular technology track, high-intensity R&D is essential to gain competitive advantage. According to the statistics of Securities Times · Databao, the R&D expenditure of multimodal AI concept stocks in 2023 will total 32.753 billion yuan, accounting for 11.2% of the revenue, which is 4.46 times the overall level of A-shares in the same period.
Yuncong Technology -UW、 Geling deep pupil 、 Alte 、 Hongruan Technology The R&D expenditure of the four concept stocks accounts for more than 50% of the revenue, which is equivalent to spending more than half of the revenue on R&D. The R&D expenditure intensity of Cloud from technology to UW is the highest. In 2023, the company's R&D expenditure will be 491 million yuan, accounting for 78.1% of the revenue. During the reporting period, the Company and Tianjin Port The PortGPT project, a large port model jointly developed by the Group, Zhejiang Harbour Group, Huawei and Tianjin Supercomputing Center, was officially launched.
Through research on listed companies, institutions can better understand their latest industrial layout. According to the statistics of Databao, a total of 61 multimodal AI concept stocks have welcomed institutional research since this year. Tom Cat 、 Suzhou Keda 、 Intesa Group 、 Zhongke Information 、 Zhongke Chuangda 、 It's worth buying The number of investigations is 5 or more. The largest number of institutions participating in the research is Zhongke Chuangda, with a total of 457; The second is Guanglianda , Tom Cat UOB Etc.
When the stock market recovers, open an account first! Intelligent fixed investment, condition sheet, individual stock radar... for you>>
Massive information, accurate interpretation, all in Sina Finance APP
