Shulou(Shulou.com)11/24 Report--

During the National Day holiday, while domestic festivals are celebrated, overseas manufacturers are busy "building a wave of big ones". Google has officially launched a new generation of Android flagship phones Pixel 8 / Pro, and announced the launch of a "Bard assistant (Assistant with Bard)" for Android and iOS devices, which allows users to interact with Bard assistants through text, voice or image-in other words, the Bard assistant launched by Google will have multimodal features.

Similarly, at the end of September, OpenAI announced that ChatGPT would introduce new voice and video features. Users can not only enter prompts in the text box, but also communicate with ChatGPT through voice or image. According to OpenAI, the new feature will be launched to ChatGPT paying users in the next two weeks and will be rolled out to other users soon.

Two well-deserved overseas AI leaders have entered a multimodal era, while the pace of domestic companies has not been slow. At present, Huawei's AI model architecture already includes the Pangu multimodal model, and the iFLYTEK Spark cognitive model launched by iFLYTEK also provides a multimodal interactive experience. In addition, Wanxing Technology (300624.SZ), a listed company of AIGC software A shares, which has been paying considerable attention to the market at the AI application level, also announced that it will soon release the "Skyscreen", the first large multimedia model focusing on video creative applications with 10 billion parameters. According to the data, the "canopy" large model also has multimodal capability.

As the technology giants and star technology enterprises gradually strengthen the support of various models for multimodal capability, "multimodal" has undoubtedly become another "hot word of the year" after AIGC and big model. This is not that big companies have a connection. In fact, multimode has been recognized by many people in the industry as an important way to general artificial intelligence (AGI).

In the early exploration of AI and deep learning algorithms, researchers mostly focus on the single modal model, and use single modal data to train the model. However, in the real world, text, image, voice, video and other forms do not exist independently, but are presented in a more complex way, just as the "five senses" of human beings are inseparable from each other. Therefore, in the exploration of artificial intelligence, cross-mode and multi-mode have become the focus of industry research in recent years.

According to industry insiders, the multimodal pre-training model integrates the processing mode of various modal information, such as voice, text, image, video, etc., which lowers the threshold of AI tasks, is closer to human perception, and has higher social value and commercial prospects, so that AI is expected to become a production tool that can be used by thousands of people.

From NLP at the beginning of the year to multimode now, stripping off the rapidly changing technology "shell", the core of the AI industry still lies in the word "application". At present, there are many domestic manufacturers in the C-end layout. From the endless ChatGPT "replacement", the wonderful duck camera out of the circle, to the popularity of the artifact of digital man's short video creation, domestic manufacturers have "blossomed in an all-round way" from text, pictures to videos, each with their own tricks to explore innovative ways of AI content generation technology, in an effort to capture the minds of domestic users.

Take Wanxing Technology as an example. As the leader of AIGC and the largest overseas company of digital creative software in China, Wanxing Technology has already started the layout of AIGC applications, and has dabbled in most of the mainstream C-end AIGC applications in the market.

Not long ago, when Wanxing Technology announced the "canopy" of the multimedia model, it demonstrated several creative software applications with the ability to integrate the large model at one time. Among them, it includes AI digital human live broadcast artifact Wanxing live broadcast version, AI text and video editing product Wondershare Kwicut, online image audio and video light editing AI creative platform Wondershare Media.io, card point music video template product Beat.ly and other audio and video AI technology application products, AI e-commerce picture generation tool Wondershare VirtuLook and other picture AI technology application products, as well as text AI technology application products such as AI lecture artifact Manxing Intelligence performance and AI virtual partner product Trumate.

In addition, public data show that creative software products such as Wanxing Meow Ying, Wondershare Filmora, Wondershare PDFelement, Yitu Map, Yitu Map, Mockitt and other creative software products of Wanxing Technology have also integrated AI capabilities, with application scenarios covering AI generation of e-commerce short videos, AI generation prototypes, AI generation flow charts / mind maps and other diagrams, AI generation text, etc.

Through the analysis of the AI product distribution of Wanxing Technology, it can be found that for domestic manufacturers, products with practical functions, such as improving work efficiency and attractiveness, may bring better benefits. As for what new players will enter the game in the future, and how will they enter the market in what areas? Everything is worth looking forward to.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.