Meta released the first "human-like" AI image creation model! List of listed companies benefiting

Source: Financial Associated Press

According to foreign media reports, Meta recently launched a computer vision model I-JEPA (Image Joint Embedding Prediction Architecture) that can learn more like humans. *It is said that the model can learn by creating an internal model of the external world * instead of making inferences based only on nearby pixels like other generative AI models.

I-JEPA demonstrates the potential of the architecture to learn off-the-shelf image representations without the need for additional assistance via human-crafted knowledge. I-JEPA uses the background knowledge of the world to fill in the missing parts of the image, and adopts the human-like reasoning method advocated by Meta's chief artificial intelligence scientist Yann LeCun, which can help avoid common mistakes in artificial intelligence-generated images , issues such as an extra finger**.

At the Beijing Zhiyuan Conference held not long ago, Yang Likun pointed out that the current GPT autoregressive model lacks the ability of planning and reasoning**, and the future GPT system may be abandoned, and he gave the correct answer in his opinion—— world model. I-JEPA is said to be the first AI model based on key components of its vision to analyze and complete unfinished images more accurately than existing models.

According to the research report released by Yan Guicheng and others on May 30, Meta's artificial intelligence has grown into an existence that cannot be ignored under the leadership of Yang Likun. *Meta is actually one of the number one AI players worthy of attention in the world *. Meta has a large number of social and metaverse algorithms, established an artificial intelligence research institute, launched the Pytorch framework, released Zion and other hardware to support the operation of algorithms, etc. Comparing Meta with Microsoft, Google, and Amazon in the field of AI, it is found that Meta is comparable to or even slightly better than the other three giants in some aspects of the field of AI.

Zhang Heng, an analyst at Guosen Securities, pointed out in a research report released on June 7 that on May 9, Meta released and open-sourced ImageBind, the first multimodal AI model that can simultaneously connect six modal data. ImageBind is committed to expanding the different modal data of the AI model, enabling AI to understand the world through the linkage of different sensory data like humans. Currently ImageBind code has been hosted on Github.

According to the incomplete arrangement of the Financial Associated Press, SuperMap Software and BlueFocus have some similar layout or business contacts with Meta.

SuperMap Software stated on the interactive platform that the company released the artificial intelligence GIS technology system and the SuperMapGIS10i product based on the technology system in 2019, which includes the ability to analyze satellite remote sensing images based on AI technology, which includes ** Semantic segmentation models similar to SAM models also include target detection AI models and scene classification AI models**. The company pays close attention to the latest developments in AI technology including Meta's new SAM model, and actively conducts research on the combination of AI technology and the company's basic software products.

Blue Cursor stated on the interactive platform that has business relations with Meta (formerly Facebook). The company has been planning and gradually developing overseas business since 2015, including technology-driven short video creation and delivery, programs Globalized media purchases, service-driven head media purchases, and one-stop overseas marketing based on platforms such as Meta (formerly Facebook), Google, TikTok, LinkedIn, Twitter, etc. AI technology has long been applied in the company.

In addition, I-JEPA and the SAM released by Meta in April belong to the same computer vision model. According to the incomplete arrangement of the Financial Associated Press, the listed companies that have recently responded to the layout of AI vision/computer vision related businesses on the interactive platform include ** Meiya Pico, ArcSoft Technology , Skyworth Digital, Runhe Software, Guokewei, Quanzhi Technology, SuperMap Software, Lingyunguang, Entrepreneurship Dark Horse, Zhongke Chuangda, Yunding Technology, Zhiwei Intelligence, Zhongke Tongda, Quectel, Opto , Kelai Electromechanical, Guolian Co., Ltd. **, etc.