All Alibaba products will access the AI foundation model "Tongyi Qianwen" in the future

Technology Author: Zixiang Li Editor: Lina Peng Apr 18, 2023 03:49 PM (GMT+8)
Alibaba

On April 11, 2023, the Alibaba Cloud Summit was held in Beijing. During the summit, Zhang Yong, Chairman and CEO of Alibaba Group, and CEO of Alibaba Cloud Computing, announced that all Alibaba products will have access to the "Tongyi Qianwen" large-scale AI model and will be providing more intelligent and personalized services. This is an important measure for Alibaba to continuously improve its artificial intelligence capabilities and user experience.

The "Tongyi Qianwen" model allows for natural language processing based on deep learning algorithms, which can understand the meaning of natural language and provide corresponding answers. By integrating this model, various Alibaba products, such as Taobao, Tmall, Alipay, Fliggy(a Chinese counterpart of Booking)and Xianyu(a Chinese counterpart of Poshmark Inc), covering e-commerce, finance, travel and other sectors, can better recognize user intentions and provide personalized recommendations and services. Users can interact through voice or text input, and receive more accurate and appropriate answers and suggestions.

After this upgrade, users can experience the following functions:

More intelligent semantic understanding: The "Tongyi Qianwen" large-scale model can deeply analyze and understand complex language information, thus more accurately grasping user needs;

More personalized service recommendations: Through analysis of user's historical consumption, search, browsing and other data, more precise service recommendations are made to users;

More smooth interaction experience: The "Tongyi Qianwen" large-scale model can better realize human-machine interaction, improving user satisfaction and usage experience.

Zhang Yong stated that in the upcoming AI era, all softwares can be rebuilt with large AI models. Based on this belief, Alibaba Cloud hopes to help more enterprises adopt large models so that they can have their own industry-specific models based on "Tongyi Qianwen". He also pointed out that developing super trillion-parameter large models is not just an algorithmic problem, but a complex systemic engineering that encompasses underlying massive computational power, networks, big data, machine learning and many other fields, requiring support from super-large-scale AI infrastructure.

Alibaba stated that by integrating the "Tongyi Qianwen" model, Alibaba products will further enhance their user experience and product competitiveness. In the future, Alibaba will continue to strengthen research and application of AI technology to bring more intelligent services to users.

With the continuous development of AI technology, the field of natural language processing (NLP) is also showing a rapid evolution trend. The emergence and iteration speed of ChatGPT have redefined the word "rapid". Prior to Alibaba, Baidu had already taken the lead in launching its generative AI "Wenxin Yiyan" based on a new generation of large language models, demonstrating its comprehensive abilities in literary creation, commercial copywriting, mathematical calculation, Chinese comprehension, and multimodal generation in five usage scenarios.

"Large models" have become a "must-win" area for internet giants. Martin Lau, Executive Director and President of Tencent, also stated that Tencent is accelerating the development of their large model "Hunyuan". On March 29th, Zhou Hongyi, founder of Qihoo 360 Technology Co. Ltd., pre-demonstrated the unreleased large model application product of 360 at an event, stating that "GPT brings about a new industrial revolution, and its significance surpasses the invention of the Internet and iPhone."

In 2020, iFlytek(CHINESE:科大讯飞) launched XLNet-Chinese, the world's largest Chinese pre-trained model with a parameter amount of 22 billion. ByteDance's AI Lab is conducting research and development of softwares similar to ChatGPT and AIGC, which may provide technical support for its VR glasses PICO in the future. Recently, Zhang Yiming, founder of ByteDance, clearly pointed out that AI will be the future strategic direction of ByteDance.

Currently, large language models are in a stage of rapid development, with vast application scenarios and commercial value yet to be fully explored. According to Meticulous Research's predictions, the global natural language processing market will grow from USD 10.22 billion in 2020 to USD 19.26 billion in 2027. The development of large language model technology will have a wide-ranging impact on fields such as machine translation, intelligent customer service, and financial risk control.

Based on current technology and market conditions, large language models are expected to develop in several directions in the future:

Firstly, higher hardware requirements will be necessary to meet the needs of increased model computing power and data volume;

Secondly, multi-lingual and cross-lingual pre-training models with stronger cross-lingual processing capabilities will need to be developed to meet the growing demand for multi-lingual information processing;

Thirdly, multi-modal integration of formats such as images, videos, and sounds into language models will be required to achieve more diverse application scenarios and accurate data analysis;

Finally, advanced techniques such as automated learning and self-supervised learning will be applied.

Future large language models will be more sophisticated, diverse, and able to meet the growing demand for data processing. At the same time, as technology develops and application scenarios expand, large language models will also become an increasingly important field, bringing more commercial value to society and enterprises.

At this critical stage, the ESG2023 "the US Sub-Forum" will be held in Shenzhen on June 2, 2023. EqualOcean will invite the representatives of academic institutions and investment institutions, as well as company leaders, from the International Studies Institute of Fudan University, Sun Yat-sen University, Jinan University School of International Relations, Gaorong Capital, Vision Plus Capital, FreeS Fund, ZWC Partners, Source Code Capital, Niu Technologies, TikTok, SHEIN, NEIWAI, miHoYo, Tuya, MINISO, Dreame, Ecovacs, WYBOTICS, Lifen, and UniUni.EqualOcean analysts will also release the US Report 2023 at the US Sub-Forum, summarizing business opportunities and best practices for expanding into the US market in the new stage. Click the link to apply and scan the QR code below to learn more about ESG2023.

飞书20230426-135940.jpg.jpg