深圳市工业和信息化局正在修订《打造人工智能先锋城市的若干措施》，对现有政策工具包进行优化升级 1。现行措施自2024年12月18日起施行，有效期为2年，今年底即将到期 1。此次修订将新增推动场景应用联合攻关和行业赋能平台建设政策，并突出支持OPC社区发展 1。
在资金支持方面，场景应用联合攻关政策每年最高投入5000万元，单个项目资助不超过研发投入的30%、最高300万元 1。行业赋能平台建设对新建平台的资助不超过审定总投入的50%、最高2000万元；已建平台则按上年度运营费用的30%给予资助，连续不超过3年、每年最高500万元 1。
此次修订的重要调整包括将原有的"模型券"调整为"词元券"，扩大资助范围至企业调用云平台部署大模型或租用智能算力部署大模型的相关费用 1。同时将"语料券"调整为"数据券"，增加对自建高质量数据集进行大模型、具身智能研发和应用的支持 1。
Shenzhen has initiated a revision of its artificial intelligence development measures to enhance its policy toolkit as existing initiatives approach expiration. 1 The current framework, which took effect on December 18, 2024, is set to expire at the end of this year after a two-year validity period. 1 The updated policy introduces new mechanisms to support joint research on application scenarios and industry-enablement platform construction, while expanding existing subsidy programs to better support enterprise innovation. 1
The revised measures introduce substantial financial commitments across multiple areas. Scenario-based joint research projects will receive up to 50 million yuan annually, with individual subsidies capped at 30 percent of project research and development costs or a maximum of 3 million yuan. 1 Industry-enablement platforms will be supported differently depending on their stage: newly built platforms can receive subsidies up to 50 percent of total approved investment costs, with a ceiling of 20 million yuan, while existing platforms qualify for 30 percent of the previous year's operational costs, continuing for no more than three consecutive years with an annual cap of 5 million yuan. 1
A key upgrade replaces the previous "model vouchers" with "token vouchers," broadening the scope of subsidized expenditures to include enterprise deployment of large language models on cloud platforms and rental of intelligent computing infrastructure for model deployment. 1 Concurrently, "corpus vouchers" have been redesignated as "data vouchers" to expand support for enterprises developing high-quality datasets for large model and embodied AI research and applications. 1
评论
还没有评论，欢迎留下第一条。