At least holding 8 large models, how can Hangzhou seize the opportunity in the AI era?

CN
巴比特
Follow
1 year ago

Source: Hangzhou Investment Promotion Bureau

On August 4th, the "AIGC and Industry Empowerment" Artificial Intelligence Industry Salon was held at the China (Hangzhou) Artificial Intelligence Town. Representatives from national artificial intelligence industry enterprises such as Alibaba Cloud, Inspur, SenseTime Technology, Tonghuashun, as well as experts, scholars, investment institutions, and representatives from relevant industry departments gathered to discuss how to actively embrace the new wave of technology and seize the development high ground in the most popular and dynamic industry direction in the industry at present, "AIGC".

Since the explosion of ChatGPT at the end of last year, AIGC (Artificial Intelligence Generated Content) has become a hot track in the AI industry, triggering a large-scale "model battle" in the industry. Overseas commercial consulting firm ARC predicts that considering the rapid increase in demand for content from the next generation of the Internet, the AIGC market size will reach 110 billion US dollars by 2030. The Quantumbit Think Tank is even more optimistic, believing that by 2030, the AIGC market size in China will reach 1.15 trillion yuan.

As the detonation point of this round of AIGC craze, large models have become a battleground for the industry. Since the beginning of this year, domestic internet giants have successively released "general large model" products, and there has been a constant stream of industry large models focusing on specific areas. Hangzhou, as the "number one city in the digital economy," has always closely followed the trend of artificial intelligence development. According to incomplete statistics, the number of large models in the country has exceeded one hundred, and Hangzhou currently has 8 large models, including Alibaba Cloud's Universal Large Model family, West Lake Heart's West Lake, Hikvision's Wutong, Hengsheng Electronics' Light-GPT, Ant Group's Zhenyi, H3C of New H3C's Baiyelingxi, Real Intelligence's Tas, and Netease Fuxi's Yuyan.

Facing the exceptionally hot AIGC track, how will Hangzhou, with its excellent internet genes, embrace this technological revolution and promote the innovative development of the artificial intelligence industry?

Large models have become a battleground for the industry, how can innovative enterprises break through

At the event, Mu Fei, Deputy Dean of Alibaba Cloud Research Institute, introduced the current progress and differences in the large model entrepreneurial market in China and the United States. He said that there are not as many choices for large models in the United States, while China is currently in a state of "a hundred-model battle." However, regardless of which large model is ultimately chosen, it actually involves a series of secondary development based on the large model and a lot of calls, in order to achieve commercialization.

Regarding the commercial application of large model products, some Hangzhou enterprises have begun to explore. At the event, representatives from Tonghuashun introduced its application of large model technology in the vertical financial field. The company has applied large model technology to intelligent dialogue robot systems, which have initially provided functions such as intelligent stock selection and stock diagnosis for investors.

Not only in the financial field, the application of large models has also accelerated the entry of artificial intelligence into more life scenarios. Peng Zhao, founder of Zhicifang, introduced that large models are increasingly reconstructing application scenarios in various fields, such as social media, search engines, and industrial robots, all of which can be seen using large models for application construction. In addition, she also analyzed the value chain of the large model industry from an investment perspective. In her view, companies focusing on related hardware, integration, and application in these areas may be more likely to achieve good results in terms of revenue and user growth.

Zhu Linchao, a hundred-person researcher and doctoral supervisor at Zhejiang University, combined the technical characteristics of "models" to analyze the application scenarios suitable for "large" and "small" models. He believes that if "large" and "small" models are organically combined, it will be beneficial to enhance the knowledge of the model, improve the applicability of the model, and achieve stronger extrapolation capabilities.

Although the opportunities and prospects brought by large models have invigorated the industry, from a technological development perspective, large model technology is still in its early stages of development. For innovative enterprises currently betting on this track, what kind of market are they actually facing?

Liu Yi, Managing Partner of Challenging Capital, stated that from the perspective of entrepreneurship in the large model track, the window of opportunity is very short, and companies need to consider which large model to base their entrepreneurship on. They not only need to explore business models but also need to find suitable binding enterprises. Additionally, as the US dollar fund is no longer the main theme of the market, and the market lacks "long-term funds," the requirements for government industry guidance funds are also higher.

Huang Wei, founder of Turing Incubator (Shanghai), stated that in some areas, domestic companies are at a similar level to overseas companies in terms of large models, but there is still a certain gap in the deep-level logical reasoning. Overall, China is still in a relatively early stage. Developing large models requires huge investment, but there is still a large development space for industry models, and this development is also constrained by computing power.

"High computing power + strong algorithm + big data," helping Hangzhou seize opportunities

Facing this new market, Hangzhou is also actively seizing opportunities. During the roundtable session of the salon, the guests discussed and provided suggestions on how Hangzhou can seize the opportunity to develop the artificial intelligence industry.

Currently, with the popularity of AIGC, there is an increasing demand for computing power support due to the processing of massive and multi-source data and large model training. Domestic intelligent computing centers and supercomputing centers are also rapidly developing, and various manufacturers have successively released new AI servers. The discussions of the guests also revolved around this market change.

In fact, Hangzhou has established many advantages in some key elements of developing artificial intelligence, such as computing power and data. According to the IDC "2022-2023 China Artificial Intelligence Computing Power Development Assessment Report," Hangzhou ranked second in the 2022 Chinese artificial intelligence city ranking. In terms of industrial ecology, Hangzhou has initiated the Ministry of Industry and Information Technology's (Hangzhou) large model evaluation platform and Zhejiang Computing Power Technology Co., Ltd. project. At the same time, Hangzhou is actively promoting the authorization and operation of public data across the city, with a total of over 3,731 open interfaces, 3,223 data sets, and 47,070 data items in the fields of credit, transportation, and medical care, with a total of over 6 billion records. In 2022, the total revenue of the artificial intelligence industry in Hangzhou reached 248.7 billion yuan, ranking fourth in the country in terms of comprehensive strength.

Based on the existing advantages, the guests conducted in-depth discussions on how Hangzhou should further develop the artificial intelligence industry. Zhu Linchao believes that at the current stage, innovation should be driven by application research and some basic research.

Miao Heng, Deputy Secretary-General of the Zhejiang Artificial Intelligence Industry Technology Alliance, believes that Hangzhou should pay more attention to the development of mid-level enterprises at this stage. Currently, the top digital economy enterprises in Hangzhou have strong strength, but the mid-level enterprises are relatively weak. Especially in the process of the emergence of leading enterprises in segmented fields, there is a lack of dedicated policy funds for attention, which can be a focus for future development.

The relevant person in charge of the Investment Promotion Bureau of Xihu District, who participated in the discussion, stated that from the perspective of grassroots investment promotion, Hangzhou should continue to make up for shortcomings based on the current needs of enterprises at the industrial base level. The needs of enterprises have changed from the original need for housing subsidies and industrial funds to the need for computing power subsidies. In the future, Xihu District will also strengthen support in this area.

During the event, the relevant person in charge of the Hangzhou Municipal Bureau of Economy and Information Technology interpreted the recently released "Implementation Opinions on Accelerating the Innovative Development of the Artificial Intelligence Industry" and detailed the relevant support measures. According to the implementation opinions, Hangzhou plans to basically form an industrial ecosystem of "high computing power + strong algorithm + big data" by 2025, realizing a virtuous cycle of large computing power incubating large models, large models driving large industries, and large industries promoting large development, making Hangzhou a national low-cost computing power area, a source of model output, and a high ground for data sharing. At the same time, the plan is to cultivate one general large model with performance reaching international advanced levels and 10 specialized models with significant industry influence by 2025.

To achieve this goal, Hangzhou will provide a series of support measures in terms of policies, technology, funds, and ecology, including encouraging leading enterprises and universities to conduct multimodal general large model research and open model applications to small and medium-sized enterprises; providing a maximum subsidy of no more than 50 million yuan for training costs for large models with a parameter size exceeding one hundred billion and performance reaching the domestic leading level; and selecting no more than 10 outstanding specialized models with advanced performance and successfully landing applications in Hangzhou each year, providing a maximum subsidy of no more than 5 million yuan.

It can be seen that, thanks to the developed digital economy, Hangzhou not only has excellent internet genes and a good industrial foundation in developing artificial intelligence, but also has many policy supports closely related to industry development, providing a fertile soil for more artificial intelligence enterprises to land in Hangzhou. With a series of elements in the development of the artificial intelligence field continuously gathering, it can be foreseen that the artificial intelligence industry in Hangzhou will also enter a new stage.

免责声明:本文章仅代表作者个人观点,不代表本平台的立场和观点。本文章仅供信息分享,不构成对任何人的任何投资建议。用户与作者之间的任何争议,与本平台无关。如网页中刊载的文章或图片涉及侵权,请提供相关的权利证明和身份证明发送邮件到support@aicoin.com,本平台相关工作人员将会进行核查。

币安:注册返10%、领$600
链接:https://accounts.suitechsui.blue/zh-CN/register?ref=FRV6ZPAF&return_to=aHR0cHM6Ly93d3cuc3VpdGVjaHN1aS5hY2FkZW15L3poLUNOL2pvaW4_cmVmPUZSVjZaUEFG
Ad
Share To
APP

X

Telegram

Facebook

Reddit

CopyLink