Shanghai Releases 11 Measures to Promote AI Large Model Development: Supports Talent Settlement in Shanghai

[baoshi.rao]

The Shanghai Municipal Commission of Economy and Informatization has released 'Several Measures to Promote the Innovative Development of AI Large Models in Shanghai (2023-2025)', aimed at fostering the innovative development of large-scale pre-trained models (large models).

The measures outline the following key initiatives: supporting large model innovation capabilities, including implementing a large model innovation support program and establishing a large model testing and evaluation center; enhancing the supply of innovation resources, including implementing a large model intelligent computing acceleration plan and building a collaborative ecosystem for intelligent chips and software.

Additionally, the measures promote innovative applications of large models, including implementing a large model demonstration application promotion plan and advancing scientific intelligent large model applications; creating a first-class innovation environment, including building a highland for large model innovation that attracts enterprises and talent, and promoting the development of a large model application ecosystem; establishing regular service mechanisms and increasing investment efforts.

These measures aim to accelerate the development of Shanghai's AI industry, build an open and secure innovation ecosystem, and create a world-class AI industrial cluster.

Several Measures to Promote the Innovative Development of AI Large Models in Shanghai (2023-2025)

To thoroughly implement the national strategic deployment for the development of next-generation AI and to enforce the 'Shanghai Regulations on Promoting the Development of the AI Industry', these measures are formulated to drive the innovative development of large-scale pre-trained models (hereinafter referred to as 'large models') in the city, build an open and secure innovation ecosystem, and accelerate the creation of a world-class AI industrial cluster.

1. Focus on Supporting Large Model Innovation Capabilities

1. Implement the Large Model Innovation Support Plan. Support the introduction of high-level innovative enterprises, assist local innovators in developing internationally competitive large models, encourage the formation of data flywheels to accelerate model iteration, and provide special rewards for significant achievements. Support local entities in conducting fundamental research on general artificial intelligence, scientific intelligence, embodied intelligence, urban large models, and other cutting-edge areas, as well as developing next-generation model architectures and training methods. Prioritize support for large model innovation within municipal-level initiatives such as strategic emerging industries, high-quality industrial development, and major scientific and technological projects.

2. Establish a Large Model Testing and Evaluation Center. Focus on aspects such as performance, safety, ethics, and compatibility to build a national-level large model testing, verification, and collaborative innovation center. Encourage innovative enterprises in the large model sector to conduct relevant testing and evaluations through this center. Support local entities in leading or participating in the formulation of national standards related to large models.

II. Enhancing the Supply Capacity of Innovation Elements

3. Implement the Large Model Intelligent Computing Acceleration Plan.

Establish a municipal-level intelligent computing power coordination and scheduling platform to build large-scale advanced computing power scheduling and supply capabilities. Eligible intelligent computing centers will receive priority support through green channels in areas such as energy consumption indicators. Accelerate the construction of key intelligent computing industry clusters in areas like Lingang, Jinshan, and Songjiang. Strengthen coordination between municipal and district levels, providing deployment rewards (after evaluation) to entities that complete intelligent computing power deployment in Shanghai and integrate it into the coordinated scheduling system by the end of 2024. For local entities renting computing power from the city’s coordinated scheduling system for large model development, subsidies of up to 10% will be granted based on cluster scale and research outcomes. Additionally, build an ultra-fast intelligent computing power network across the city, enabling 100 Gb/s+ high-speed direct connections and millisecond-level latency between intelligent computing centers.

4. Build a smart chip hardware-software collaborative ecosystem. To support the development and application of large models, the city will assist local smart chip companies in scaling their applications and conducting validations. Efforts will be made to establish a smart chip hardware-software adaptation system to reduce adaptation costs for enterprises. A smart chip and hardware-software adaptation evaluation center will be established in Shanghai. Eligible hardware-software adaptation products will be included in the support scope for first-batch and first-edition initiatives.

5. Shared construction of corpus data resources. A large model corpus data alliance will be formed to encourage diverse stakeholders to jointly promote the development of high-quality corpus data elements. A framework for intellectual property protection of corpus data will be established, leveraging technologies like blockchain to deepen contribution-driven incentive and evaluation mechanisms. A corpus data trading platform will be set up under the Shanghai Data Exchange.

III. Promoting Large Model Innovation and Application

6. Implement the Large Model Demonstration Application Promotion Plan. Focus on supporting the construction of demonstration application scenarios in fields such as intelligent manufacturing, biomedicine, integrated circuits, intelligent education, fintech, design creativity, autonomous driving, robotics, and digital government, aiming to create benchmark large model products and services. Promote the application of large models in empowering industrial internet platforms. Include eligible large model applications in the artificial intelligence demonstration application list and the innovative product recommendation catalog. Support local state-owned enterprises and institutions in opening up large model application scenarios, prioritizing the adoption of tested and evaluated large model products and services.

7. Advance the Application of Scientific Intelligent Large Models. Support relevant entities in establishing platforms such as scientific intelligent innovation centers and algorithm innovation bases, coordinating computing resources and scientific datasets, and promoting the application of scientific intelligent large models in fields such as life sciences, engineering calculations, and meteorology, creating new paradigms for scientific research.

IV. Creating a First-Class Innovation Environment

8. Build an innovation hub for large models to attract enterprises and talents. Encourage Pudong New Area, Xuhui District, and others to establish large model ecosystem clusters, focusing on R&D and industrialization of large models, and increase support efforts. Encourage the Lingang New Area of the Free Trade Zone to explore international exchanges and cooperation related to large models. Prioritize the inclusion of key talents in large model innovation into national and municipal high-level talent programs, with a focus on supporting the settlement of skilled professionals in large model-related fields. Organize joint training of interdisciplinary large model talents by enterprises, universities, and research institutions.

9. Promote the construction of a large model application ecosystem. Create an open-source large model industry application innovation ecosystem, support the development of large model open-source communities and collaboration platforms, and attract and cultivate startup teams related to large models. Strengthen the deep application of industry-specific large models in vertical fields and build an open and collaborative industrial ecosystem.

10. Establish a regular service mechanism. For internet information services driven by large models that have public opinion attributes or social mobilization capabilities, provide regular contact services, strengthen compliance guidance, and promote relevant entities to fulfill procedures such as security assessments and algorithm filings in accordance with regulations. Explore innovative regulatory mechanisms within relevant clusters.

11. Increase investment efforts. Comprehensively promote the development of large models and related fields, further strengthen the artificial intelligence industry fund, better guide social capital participation, and increase resource investment in the city's artificial intelligence industry.

This notice shall take effect from the date of issuance.