Large model group competition better empowers thousands of industries

Lu Liyuan, reporter of Science and Technology Daily, Wang ChunIn the past two days, there have been a lot of people in the Shanghai World Expo Exhibition Hall. Digital human anchors, Chatbot, AI postcards, AI painting and poetry generating videos

Lu Liyuan, reporter of Science and Technology Daily, Wang Chun

In the past two days, there have been a lot of people in the Shanghai World Expo Exhibition Hall. Digital human anchors, Chatbot, AI postcards, AI painting and poetry generating videos... These applications let people really feel that AI technology is gradually infiltrating into our lives. Help me create a travel guide to Yunnan, write a poem praising my wife, make me a cup of coffee, etc. In the exhibition hall, people have put forward various requirements for various applications on the booth and received accurate responses from artificial intelligence. The interaction between humans and artificial intelligence has become more diverse and smooth.

People in front of the SenseTime booth are experiencing the application of large models. Photographed by Science and Technology Daily reporter Wang Chun

During the World Artificial Intelligence Conference (WAIC) held from July 6th to 8th, the topic of large models was undoubtedly the hottest focus of this conference. More than 30 domestic and foreign large models were displayed, and almost all of the top products in the vertical field of basic application large models were present. Therefore, Hu Houkun, the rotating chairman of Huawei, couldn't help but sigh that the biggest difference between this year's conference and previous ones is the emergence of ChatGPT, which has pushed artificial intelligence to a new level.

Accelerating the Application of Artificial Intelligence Industry with Large Models

At this year's WAIC conference, generative AI has become a new field of competition for various internet technology companies. "With the help of the Spark Model, iFLYTEK AI learning machine corrects compositions like a teacher, conducts live dialogue, accompanies training like a speaking teacher, and conducts real-time oral error correction..." At the IFlytek booth, staff demonstrated the application results of iFLYTEK's Spark Model in the educational scene. "IFLYTEK is relying on an autonomous and controllable large model training base to promote the continuous evolution of the large model, so that the large model can enable thousands of industries," said the relevant person in charge of IFlytek.

In March this year, as one of the representative enterprises in the domestic AI field, Baidu released the big language model "ERNIE Bot", becoming the first ChatGPT like product in China. At this AI conference, Baidu brought the same AIGC model as ERNIE Bot Wenxin, which can not only inspire visual Content creation such as painters and designers, assist in artistic creation, but also provide high-quality and efficient pictures for Content creation such as media and authors, and is regarded as one of the "nine treasures of town hall" of this conference.

The industry's first AI protein generation model, "Darwin," was officially unveiled on July 7th at WAIC. This big model can realize multi-mode directional generation by learning hundreds of billions of levels of multi-mode Big data. A single model can meet the needs of the whole process of protein generation, such as sequence generation, structure prediction, function prediction, and de novo design, solve the problem of specific functional protein generation required by industrial applications, and evaluate the effect and value in the real industrial environment. Xu Jinbo, the model R&D leader, an outstanding visiting professor of the Institute of Intelligent Industry of Tsinghua University and the founder of Molecular Heart, said that the AI protein generation model aims at the real industrial application needs such as innovative drug design and Synthetic biology, and will use a model to meet the needs of the whole process of protein generation. In the future, protein designs such as macromolecular drugs and new biomaterials can achieve "one click customization".

People are experiencing big model applications in front of the Baidu booth. Photographed by Science and Technology Daily reporter Wang Chun

With the emergence of the "Hundred Model War" in China, vertical industry large models will also face more severe competition. At the meeting, Mido released its first knowledge Q&A and content generation big language model that supports localized software and hardware operating environments - Mido. It is reported that this model is built on the basis of high-quality Chinese multimodal data training of hundreds of billions, which can not only achieve customized content generation of "thousands of texts, thousands of people, thousands of faces", but also create an "exclusive" large model for the government and enterprises based on the Huawei Shengteng Shengsi AI ecosystem.

The research on universal big models is aimed at enabling enterprises without the ability to build models to enjoy the convenience brought by big models, which is more suitable for top enterprises with stronger capabilities to do. Industry leading companies are more suitable to focus on the industry and create vertical industry big models. "Liu Yidong, the head technical officer of Mido, said that currently, the Mido series of industry big models have gradually attempted to be deployed in scenarios with strong demand for content production such as government affairs and media.

Developing large models still requires consolidating the underlying foundation

With the coexistence of hundreds of "models" and the competition of various schools of thought, the artificial intelligence industry needs to develop step by step, steadily, and solidly, to consolidate the underlying foundation.

At the WAIC conference, companies such as Baidu, Huawei, and Alibaba were awarded the leadership unit of the "National Artificial Intelligence Standardization Overall Group Large Model Special Group" by the China Electronics Technology Standardization Research Institute of the Ministry of Industry and Information Technology, actively promoting the construction of a national standard system for large models and assisting the development of China's large model industry.

On July 7, during the WAIC Conference, Alibaba and SenseTime launched the latest version of the big model. Alibaba Cloud announced a new AI painting creation model, which extends the model from text and voice to image, and gradually moves closer to multimodal models. This model can assist humans in image creation and can be applied in future application scenarios such as art design, e-commerce, games, and cultural creativity.

It is reported that Tongyi Wanxiang is the product of Ali's years of hard training in the field of big models. It is based on the composite of the combined Generative model developed by Ali. At last year's World AI Conference, Alibaba Cloud Cloud's "Tongyi" model series debuted for the first time. After that, models and products such as Tongyi Qianwen, Tongyi Listening and Understanding were launched successively. At present, Alibaba Cloud's big model is gradually becoming more and more popular. More than 300000 enterprises have applied for testing Tongyi Qianwen, and the cumulative number of users of Tongyi has reached 360000. "Multimodal is the inevitable path for the evolution of the big model. We hope to use different modal capabilities to serve thousands of industries and help enterprises fully embrace the intelligent era." Zhou Jingren, CTO of Alibaba Cloud Intelligent Group, said.

Alibaba Cloud AI painting creation model was unveiled. (Courtesy of the organizer of the conference)

SenseTime has also upgraded the "Shangtang RiRiXin SenseNova" model on WAIC, which has not only greatly improved the model parameters, but also made further breakthroughs in multimodal capability. Shang Tang stated that he hopes to promote the leapfrog improvement of AI infrastructure capabilities through "large models+large devices".

The breakthrough of large models has sparked a new round of technological revolution in artificial intelligence, followed by explosive growth in industrial demand, and new application scenarios and models are rapidly emerging. Xu Li, chairman and CEO of SenseTime, said at the AI forum: "Shangtang will not only build a basic model with more powerful general capabilities, but also further integrate professional knowledge in different vertical fields, build a professional model with more industry knowledge and expertise, and fundamentally reduce the cost and threshold of downstream applications of the big model."

Disclaimer: The content of this article is sourced from the internet. The copyright of the text, images, and other materials belongs to the original author. The platform reprints the materials for the purpose of conveying more information. The content of the article is for reference and learning only, and should not be used for commercial purposes. If it infringes on your legitimate rights and interests, please contact us promptly and we will handle it as soon as possible! We respect copyright and are committed to protecting it. Thank you for sharing.(Email:[email protected])