Integration of cloud and intelligence to jointly build a new ecology in the era of large models

afb6c06576052fbfe9024f0ce5aa90ac.gif

On May 18, Baidu Smart Cloud held the 2023 Baidu Smart Cloud Partner Conference in Guangzhou. At the meeting, 28 ecological partners were signed. On the one hand, they will jointly explore the innovative application of the Wenxin Qianfan large-scale model platform in various fields, and on the other hand, they will work together to create industrial benchmark cases and replicate them in a standardized and large-scale manner in various industries.

Shen Dou, Executive Vice President of Baidu Group and President of Baidu Smart Cloud Business Group, said that Baidu Smart Cloud is willing to work with partners to maximize the opportunities brought by generative AI and large models, and jointly promote the intelligentization of Chinese society to a higher level.

Yuan Foyu, vice president of Baidu Group, said that the era of large-scale models will bring three major industrial opportunities, new cloud computing, fine-tuning of industry models, and application development. No matter which layer of the three layers customers and partners want to invest in, Baidu Smart Cloud can provide the reliable services they need, including the world's first one-stop enterprise-level large-scale model production platform Wenxin Qianfan, and the full-stack self-developed AI era infrastructure Baidu AI Big Base.

The circle of friends in the smart cloud ecosystem continues to expand

After years of rapid growth, Baidu Smart Cloud has given full play to the core advantages of AI and built a complete ecological partner system in the sales, technology, service, and cloud markets. Since the internal testing of Baidu Wenxin Yiyan, the construction of a new ecology around large models has been steadily advancing. At present, more than 300 partners have participated in the internal test of Wenxin Qianfan large-scale model platform, and achieved test results in more than 400 internal scenarios of enterprises. Baidu's financial report for the first quarter of 2023 shows that Baidu Smart Cloud achieved profitability for the first time in this quarter. Sales leads increased by more than 400% year-on-year in March, and revenue was 4.2 billion yuan, an increase of 8% year-on-year, exceeding market expectations.

At the meeting, Baidu Smart Cloud signed contracts with 10 Wenxin Qianfan ecological partners. In the future, in the era of large-scale models, it will comprehensively empower partners from technical services to industry applications, and work together to create one-stop products, solutions and services for digital transformation and intelligent upgrades, making industrial intelligence easier.

bc803571b11578dcac3b68edc7aa1fb5.jpegWenxin Qianfan ecological partner signed a contract

d15aecb3286b29176853123f39212c38.jpeg

21be4f6e335f9435f782cf9d11bfb17e.jpeg

e7688322177b7fd853095c3f60e2941c.jpeg

ef41e9374b4a32521548af44d01a9920.jpeg

abdc84d842613b31465d3a7dcf04446b.jpeg

a948a1c2bf192ad03543e8c6cc07e109.jpeg

def3ae510aa0ef57260b067b0168ccc5.jpeg

6d98d6c1f68656a3432de315fabc5a9a.jpeg

37043fe67bd4bd56bd6bdcc140d633c8.jpeg

outside_default.png

Swipe left and right to see more

outside_default.png

Baidu Smart Cloud also cooperates with upstream and downstream partners in the industry to jointly create a new industrial ecology of "integration of cloud and intelligence", is committed to the construction of standardized and large-scale artificial intelligence solutions, and accelerates the integration of the digital economy and the real economy. At the meeting, 18 regional partners reached a strategic cooperation with Baidu Smart Cloud to boost the digital and intelligent transformation of the industry.

dd7b28b7431759fe5ea5b656193c8521.jpeg

Cloud-Intelligence Integrated Industry Alliance was established

47bcf70560eefc6a381972ccc7706c0c.jpeg

6650972d077c9275b3f1dedf1b8bb04f.jpeg

a159ba91bca6b2af9735fa2ed30075ca.jpeg

79960a6cd2f544c134c172ce503c74b2.jpeg

cea297d2ef92b8cb4d8e2ef79f469a69.jpeg

130b74b860623f2d8572b44ca96a65fd.jpeg

c4bae0fe668221d3eba51504341df944.jpeg

5bde0e73aa587e4721e3a19d49420c5c.jpeg

50ffa8632dee509095339e87edaa9f81.jpeg

c963ee289cbace1db998fbbea68a6c52.jpeg

72aa0bf7f0df942969ad0289855023a0.jpeg

440e41f4d18dff7c6f2cbdede4210837.jpeg

5e069318a48e5904e0041d749e24c2c0.jpeg

3df8ae08a3dfabf98a9bd6df868303e6.jpeg

8004dbf8af83924aea9e5694fc9d5348.jpeg

7d633f6f73d05e30d867f53a0dea512e.jpeg

35f371afb77eb8995ec8622dfaf3a9a8.jpeg

f6d9f752149cf3a374b0e31dfdd5b3b1.jpeg

outside_default.png

Swipe left and right to see more

outside_default.png

Baidu Smart Cloud adheres to the basic principle of "industry deep cultivation and continuous sinking of cloud-intelligence integrated model innovation ecological construction AI base construction". At present, it has completed the layout of ecological partners in cities, water affairs, industry, education, pan-technology and other industries in Guangdong, Guangxi, Fujian, Hainan, Yunnan, Chongqing and other places to jointly create end-to-end joint solutions and realize large-scale replication on mature tracks. As Meng Fanying, deputy director of the Guangzhou Baiyun District Water Affairs Bureau, said, with the help of Baidu's AI capabilities, 24-hour inspections and supervision of water conservancy facilities, water work sites, storage yards, culverts, etc., have become a breakthrough point for the Baiyun District Water Affairs Bureau to reduce costs and increase efficiency.

The best environment for enterprise R&D and application of large models

In order to efficiently help customers and partners seize the industrial opportunities in the era of large-scale models, the Wenxin Qianfan large-scale model platform created by Baidu Smart Cloud provides large-scale model services including Wenxin Yiyan for enterprises, and also supports third-party large-scale models, provides various tool chains and a complete set of environments for developing AI applications, and provides enterprises with the best environment for using, developing or applying large-scale models, allowing enterprises and developers to use and make good use of large-scale models in the simplest way.

Currently, Wenxin Qianfan large-scale model platform supports the most and most flexible deployment methods in the industry. In terms of public cloud services, three services will be provided: reasoning, fine-tuning, and hosting; in terms of privatized deployment, software authorization, integration of software and hardware, and leasing services will be supported. It should be emphasized that the integrated software and hardware solution also supports various types of chips to meet the diverse needs of customers.

At the scene, the eco-partner representative Hande Information demonstrated the exploration results of four intelligent scenarios based on the Wenxin model: smart shopping guide, smart logistics, asset assistant, and marketing creative assistant. Huang Geng, chief operating officer of Hande Information, said that after this period of in-depth joint testing with Wenxin Yiyan, it can be confirmed that in China, Hande’s choice is Wenxin Yiyan, and all of Hande’s independent products have been integrated into Wenxin Yiyan’s capabilities.

Han Geying, Assistant President of Kingdee and General Manager of the Product Management and Marketing Operations Department, shared her practice and thinking on intelligent management in the era of large-scale models, and pointed out that Kingdee has already connected to the Wenxin Yiyan large-scale model and is constantly exploring and verifying in various business scenarios . In the future, Kingdee will continue to upgrade Kingdee's original NLP, OCR and other applications in combination with large models.

In addition, Wenxin Yiyan has been successfully tested internally in Baidu and in the business scenarios of hundreds of customers and partners. In Baidu, key product indicators such as accuracy, recall, and satisfaction have been greatly improved by accessing large-scale search, short videos, maps, and Xiaodu. In terms of cooperation with customers and partners, Baitu Biotech has trained a large cross-modal model with 100 billion parameters on Baidu Smart Cloud, and contributed a number of research and development cases of high-performance warheads and new functional sensors; Kingsoft Office has started joint tests in multi-modal generation such as intent understanding, model essay writing, and text-based diagram generation. In the future, cooperation is expected to be carried out at more functional and product levels.

Adapt to the new infrastructure of the era of large models

While empowering thousands of industries, Wenxinyiyan is also constantly "self-evolving". After more than a month of internal testing, it has completed 4 major technical upgrades, and the performance of large-scale model reasoning has increased by nearly 10 times. Fast and convergent is a very important indicator in large model application training. Wenxin Qianfan large model platform can reach a state of convergence faster in terms of multi-machine multi-card training performance, ranking first in the world in the world's authoritative AI benchmark evaluation MLPerf list.

Such outstanding performance is due to the six core advantages of the Wenxin Qianfan large-scale model platform, which is easier to use, more comprehensive, more efficient, safer, more open, and more integrated. It provides enterprises with a tool chain covering the entire life cycle and various plug-ins that expand the boundaries of model capabilities, helping enterprises achieve end-to-end full-process optimization, support out-of-the-box use, and save opportunity costs.

The reason why Wenxin Qianfan is capable of controlling large-scale heterogeneous computing power and providing high-level and easy-to-use large model services is because Baidu AI Big Base is suitable for a wider range of AI scenarios and scientific computing scenarios, including building a proprietary model to solve a specific problem. With the support of Baidu AI Big Base, the large model can be efficiently and stably trained and effectively converged, which can increase the kilocalorie parallel acceleration ratio to more than 90%, the resource utilization rate of the training scene is more than 70%, and the iteration efficiency of model development is increased by more than 100%.

Share the technical dividend of the era of large models‍‍‍‍‍‍‍‍‍‍‍‍

Baidu's dedication to innovation is consistent, and its investment in technology is long-term and has never wavered. We have invested in the field of artificial intelligence for 13 years, and the research and development investment has reached as high as 100 billion. Two years ago, I realized that artificial intelligence technology had undergone a directional change, from discriminative artificial intelligence to generative artificial intelligence. Since then, AIGC has become a direction of focus, and resources are gradually tilted in this direction.

Today, the industrial opportunity brought by the large model confirms our investment, which can be called another iPhone moment. Baidu Smart Cloud is willing to share the technological dividends of the large-scale model era with the industry and embark on a new journey of industrial intelligence.

d15f3e131337625ec7c51a28849a31d6.png

8bb2f5f0eb10e6e8ba1deff39936a8a1.png

edb25c9e0b24ce0bda747ac9f4042c1a.png

d7db8d42631999e4e637973bcda932c8.png

7f03ec7cb20378ccd9f43d92212201ae.jpeg

Click "Read the original text" to cooperate and consult immediately

Guess you like

Origin blog.csdn.net/weixin_48493350/article/details/130758193