Multimodal models are facing multiple benefits, and the main forces have already taken action.

The A-share market rebounded collectively today, with the Shanghai Composite Index rising by 0.69%, the Shenzhen Component Index and the ChiNext Index rising by 1.19% and 0.85% respectively.

The consumer goods sector strengthened collectively, with stocks of Baijiu companies like Huangtai Liquor and Jiugui Liquor soaring, and food stocks like Ganyuan Food, Qiaqia Food, and Qianwei Yangchu hitting the daily limit; the concept of state-owned enterprise reform continues to be active.

The insurance and China Shipbuilding Industry concepts fell against the trend.

Wanlian Securities stated that the entire food and beverage industry still has strong resilience, coupled with the general decline in upstream raw material prices, and the performance of the food and beverage industry is expected to maintain steady growth.

Advertisement

As inventory gradually clears and consumption continues to recover, Baijiu companies that have experienced a valuation kill but still have a good fundamental, as well as mass food with strong demand rigidity, are expected to usher in a valuation repair market.

Multiple benefits are coming, and multimodal large models are expected to bring an application explosion.

The multimodal model index rose by 3.38% today, ranking at the forefront of the A-share concept index increase, among which Sound Information's stock hit the daily limit, and the increases of Dang Hong Technology, Gravity Media, and New Continent were at the forefront.

In terms of news, the multimodal model field has ushered in multiple benefits.

On September 19, Alibaba International AI team released the multimodal large model Ovis, which can be applied to mathematical reasoning Q&A, object recognition, text extraction, and complex task decision-making.

Currently, the data, models, training, and inference code of Ovis 1.0 and 1.5 have all been open-sourced and reproducible; the weight of Ovis1.6-Gemma2-9B in the Ovis1.6 series has also been open-sourced.

In addition, the Institute of Geography and Resources of the Chinese Academy of Sciences held a press conference in Beijing today, officially releasing the world's first multimodal geospatial science large model "Kunyuan" (Sigma Geography), aiming to promote the deep integration of geography and artificial intelligence.

Currently, "Kunyuan" has supported the publication of more than 10 high-level academic papers such as Nature's sub-journals, The Innovation, and Earth's Future.

In the future, the development team will promote the development of map large models and geographic reasoning machines, hoping to enable the geospatial language large model to understand maps; it will also create a large platform for geographic scientific research collaboration, hoping that every scientist and research team can have their own geospatial large model, and can collaborate with millions of scientists through sharing data, models, research ideas, and other methods.

According to Su Fenzhen, Deputy Director and Researcher of the Institute of Geography and Resources of the Chinese Academy of Sciences, "Kunyuan" is currently in the trial stage.

After the completion of the three stages of development, it will shift from supporting scientific research to supporting the industry.

It can also become a geography teacher and travel guide for ordinary users.

Western Securities believes that interactive multimodal large models are another important milestone in the development of AI, breaking the limitations of traditional models on a single data type, and creating a smarter interaction method.

The ultimate form of the large model is to evolve human-computer interaction to the most primitive and simplest form.

Perhaps the main interaction method in the future will be to speak directly with computers and mobile phones.

The simplification of the interaction mode will greatly reduce the threshold for using AI, so the more intuitive and foolproof interaction brought by the interactive multimodal large model is expected to bring a wider area of popularization of large model applications.

According to the statistics of Securities Times · Data Treasure, among the multimodal concept stocks, iFLYTEK, Qihoo 360, and Kunlun are at the forefront in terms of market value scale.

Among them, iFLYTEK's latest market value is 79.94 billion yuan, ranking first.

The company launched a robot development platform with AI capabilities such as Nebula, multimodal interaction, intelligent motion, model training, asset generation, and hardware and software access, and released the iFLYTEK Xinghuo Cognitive Large Model V2.0.

Since September, Zhenshutong, Tom Cat, Cloud Cong Technology-UW, Wanxing Technology, and Yun Ding Technology have all been surveyed by more than 10 institutions.

Zhenshutong said in the survey that the company is committed to empowering the multimedia video industry with artificial intelligence technology, truly realizing the integration of audio-visual + AI.

The company's R&D team applying artificial intelligence is developing a video conference intelligent voice control box for the multimedia business.

This box empowers video conferences with AI voice.

Users can give control commands through a microphone using natural language.

It can understand the user's operational intentions and translate them into operation commands for specific devices, greatly improving the experience and efficiency of video conference management.

Tom Cat said in a recent survey activity that the company's multimodal model "Tom Cat Emotional Companion Vertical Model" jointly created with Xihu Xinchen has completed the construction and training of the main functions, and has completed the entire end-to-end link through the combination of software and hardware.

After being officially put into use, the Tom Cat vertical model will serve as the "electronic brain" of the company's "chatting Tom Cat" AI product line, empowering different terminal forms of AI applications and hardware products.

Looking at the performance expectations for the whole year, among the multimodal model concept stocks that have been rated by more than 5 institutions, the annual net profit of Tuoersi, Yi Dian Tian Xia, New Continent, Wanxing Technology, and Chinese Online is expected to achieve a significant increase.

From the perspective of capital, iFLYTEK, Chinese Online, Qihoo 360, Jiecheng Shares, Tom Cat, and Shengxun Shares all received net inflows of main funds exceeding ten million yuan today.