Ant Group WAIC releases large model encryption platform to help large models solve data supply challenges

For large models to take root in the industry, we must solve the challenge of high-quality data supply. On July 5, the 2024 World Artificial Intelligence Conference entered its second day. As a major technical service provider in the field of data elements, Ant Group released the "Language Cloud" large model encryption platform. Through trusted privacy computing technology that combines software and hardware, Achieve dense data flow in large model hosting and large model inference to protect model assets, data security and user privacy. At present, high-quality data supply and safe circulation have become the primary challenges for the application of large models in vertical industries. First, in order for large industry models to gain the ability to solve professional problems, they must first be trained with sufficient and high-quality professional data. However, professional data is often scattered among different institutions and enterprises, and is difficult to flow due to its high value and high confidentiality requirements. In addition, there are trust barriers between enterprises, large model manufacturers and users. Enterprises are worried about data leakage, large model manufacturers are worried about the security of model assets, and users are worried about personal data and privacy risks.

At the press conference, Wang Lei, CEO of Ant Computing, introduced that the "Lingyu Cloud" large model encryption platform launched the first two major services, which are:

1. Large model dense state hosting: protecting model assets from being leaked and Misappropriation.
2. Large model dense state reasoning: protect data security and business confidentiality during user interaction.

In addition, the platform supports:

• GPU calculations under a trusted execution environment.
• Memory, disk encryption and other methods to achieve end-to-end encryption and cross-domain access control.
• Lightweight remote authentication.

At present, privatized deployment of large models not only increases operation and maintenance costs, but also affects service efficiency and quality. The Lingu Cloud large model computing platform provides public cloud and private cloud delivery solutions and supports general large models.

Wang Lei revealed that the platform services will gradually cover the full link of vertical large model data security, including:

• Dense state computing services.
• Full-link tool for developing large dense state models.

Wei Tao, Vice President and Chief Technology Security Officer of Ant Group and Chairman of Ant Computing, believes that privacy computing technology determines the upper limit of cross-domain data supply. Ant Group began exploring privacy computing technology in 2016, hoping to work with industry partners to promote industrial development.

At present, the core technology of Ant Trusted Privacy Computing has been open sourced, including the "Language" framework and the "Star Blossom" operating system. In addition, Ant has also worked closely with universities and established a "Special Research Fund for Privacy Computing".

In May this year, Ant Group established Zhejiang Ant Computing Technology Co., Ltd., a dense computing company, which will provide end-to-end data security, computing acceleration solutions combining software and hardware, and a privacy computing cloud service platform.

The above is the detailed content of Ant Group WAIC releases large model encryption platform to help large models solve data supply challenges. For more information, please follow other related articles on the PHP Chinese website!