在2024年世界人工智能大会（WAIC）的“云帆奖”论坛上，蚂蚁集团遥感大模型负责人王剑带来了主题为“多模态遥感大模型的探索与实践”的深度解读。这一活动由上海人工智能实验室、机器之心、全球高校人工智能学术联盟主办，得到了世界人工智能大会组委会办公室及上海市徐汇区人民政府的指导。论坛汇聚了来自全球顶尖学府和机构的30余位云帆奖得主，共同探讨AI前沿，为中国AI发展贡献新生力量。
王剑在演讲中总结了遥感大模型的当前发展状态和未来机遇，分享了蚂蚁集团在这一领域的最新进展。蚂蚁集团基于其自主研发的“蚂蚁百灵大模型平台”，推出了一款名为“SkySense”的20亿参数多模态遥感模型。SkySense通过在数据处理、模型架构和无监督预训练算法等方面的创新，展现出卓越性能，在包括土地利用监测、地物变化检测在内的7种常见遥感感知任务中均排名第一，且在17项评测中均取得优异成绩。
王剑进一步介绍了SkySense在实际场景中的应用案例。在农村金融领域，SkySense通过精准的遥感数据，为农村信贷提供决策支持，有效提升了金融服务的覆盖面和效率。在蚂蚁森林项目中，SkySense则应用于林地保护，通过智能监测系统实时预警森林火灾和非法砍伐行为，为生态保护和可持续发展提供了强有力的技术支撑。
### 结语
蚂蚁集团通过SkySense的探索与实践，不仅展示了多模态遥感大模型在技术层面的创新成果，更在实际应用中体现了其对社会经济和环境的积极影响。这一案例不仅为中国AI产业的发展提供了宝贵的经验，也为全球AI领域的创新实践树立了新的标杆。随着AI技术的不断进步和应用的不断深入，我们可以期待更多类似SkySense的创新成果，为人类社会带来更加智能、高效和可持续的发展路径。
英语如下：
### Exploring and Applying Multimodal Remote Sensing Mega Models: Ant Group’s Innovation and Practice
At the “Cloud Sail Award” forum during the 2024 World Artificial Intelligence Conference (WAIC), Wang Jian, the head of Ant Group’s remote sensing mega model, presented a deep dive into “Exploring and Applying Multimodal Remote Sensing Mega Models”. This event was co-organized by the Shanghai Artificial Intelligence Laboratory, Machine Intelligence, and the Global University AI Academic Alliance, and was guided by the WAIC Organizing Committee Office and the Shanghai Xuhui District Government. The forum gathered over 30 Cloud Sail Award winners from top global universities and institutions to discuss AI frontiers and contribute to the growth of China’s AI sector.
Wang Jian summarized the current state and future opportunities of remote sensing mega models, sharing Ant Group’s latest advancements in this field. Ant Group, leveraging its self-developed “Ant Bailing Mega Model Platform”, launched a 20-billion-parameter multimodal remote sensing model named “SkySense”. SkySense demonstrated outstanding performance through innovations in data processing, model architecture, and unsupervised pre-training algorithms, ranking first in seven common remote sensing perception tasks such as land use monitoring and object change detection, and achieving excellent results in 17 evaluations.
Wang Jian also introduced practical applications of SkySense. In the rural finance sector, SkySense provided decision support with precise remote sensing data, effectively expanding the coverage and efficiency of financial services. In the Ant Forest project, SkySense was applied to forest conservation, using an intelligent monitoring system to provide real-time alerts for forest fires and illegal logging, providing robust technological support for environmental protection and sustainable development.
### Conclusion
Through the exploration and practice with SkySense, Ant Group not only showcased the technological achievements of multimodal remote sensing mega models, but also demonstrated their positive impact on society, economy, and the environment in practical applications. This case not only provides valuable experience for the development of China’s AI industry but also sets a new benchmark for innovative practices in the global AI field. As AI technology continues to advance and applications deepen, we can look forward to more similar innovative results from SkySense, leading to more intelligent, efficient, and sustainable development paths for human society.
【来源】https://www.jiqizhixin.com/articles/2024-07-12
Views: 1