Ant sets up a shared intelligence alliance to break the "data island"

With the development and maturity of federated learning and trusted computing technology, technology giants have begun to develop technology alliances and build ecosystems in the era of data sharing.

This month, after Tencent released the Trusted Computing Platform (TEE) at the Global Digital Ecology Conference and Huawei's Release of the ModelArts3.0 platform that supports federated learning at the Connectivity Conference, Ant Group also established the "Shared Intelligent Technology Alliance" for the first time at the Bund Conference in Shanghai from September 24 to 26.

An interconnected network based on privacy protection

It is reported that there are nearly 30 start-up members of the Ant Group Sharing Intelligent Technology Alliance, including financial institutions such as Everbright Bank, MYbank, ZhongAn Insurance, Ant Group, Alibaba, Lenovo, Inspur, Unicom Big Data, Nvidia, Tree of Light, Hammer Weiwei Technology, Wanda Information, Spichi and other technology companies, as well as China Electronics Technology Standardization Research Institute, Zhejiang University, Shanghai Jiaotong University, Institute of Information Technology of the Chinese Academy of Sciences, Zhijiang Laboratory, University of International Business and Economics and other scientific research institutions.

At present, shared intelligence has been applied in business scenarios such as intelligent marketing, intelligent risk control, and joint medical care within Ant and its partners. The establishment of the alliance will aggregate and share industry resources, promote the technology openness of shared intelligence, and maximize the value of collaboration between institutions under the premise of security and compliance.

In the financial field, there are a large number of corresponding financial scenarios such as micro-loans, insurance, payments, and wealth, which require the use of cross-institutional collaboration to reduce risks and improve efficiency. However, due to privacy and compliance reasons, data islands have been formed between various institutions, and intelligent capabilities can only be limited to isolated islands. Creating an open platform and enabling intelligent capabilities to cross silos and create greater value has become an urgent need for many institutions.

Zhang Jiachen, a participant and technology provider of Ant Group's Shared Intelligence Alliance and the founder and CEO of Tree of Light Technology, a tencent cloud privacy computing technology cooperation company, said in an interview with First Financial Reporter during the Bund Conference: "Technology companies, including giants such as Tencent and Ant, are enthusiastically embracing federal learning and trusted computing. Previously, a number of alliances have been established in the industry, but mainly alliances based on a single technology, and the alliance initiated by ants this time covers a number of technologies and upstream and downstream ecological partners that integrate the industrial chain. This is the biggest difference from the past. ”

Zhang Jiachen told the first financial reporter that the core support of China's strategy of data production factors is new technologies and new ecological business models, not a problem that can be solved by a simple technology. This is the biggest significance of the establishment of the Ant Group Shared Intelligence Alliance. "How to make data safe and unobtrusive for the application of artificial intelligence requires the ability to use (adapt) a combination of technologies that can be used (adapted) to different scenarios." Zhang Jiachen said, "These technologies cover trusted computing environments (i.e., confidential computing), federated learning, secure multi-party computing, differential privacy, and more. ”

Trusted Computing and Federated Learning are in the spotlight

Trusted computing and federated learning are currently the most talked about blockbuster technologies in the world. Yang Qiang, chief AI officer of WeBank and chair professor of the Hong Kong University of Science and Technology, introduced to the first financial reporter that the main purpose of the so-called federal learning is to keep the data in place, and the data can be used by different institutions in the encrypted state, but all parties cannot see each other's data, which can be summarized as "the data does not move the model, and the data is available and invisible."

Fan Lixin, chief scientist of the artificial intelligence department of WeBank, told the first financial reporter: "The difference between ants' shared machine leanring technology and federated machine learning technology is that the federated machine learning technology does not require data to be passed out from the original institution, and shared learning is to transmit encrypted data and operate in a trusted environment." ”

Fan Lixin stressed to the first financial reporter that this requires this central node to achieve "credibility" in two aspects. "One is legal compliance arrangements based on the law, and the other is technical credibility, such as how to defend against the outside world that may crack data in a trusted environment through some technical means." he said.

According to reports, the current technical framework of federated learning has been formed, reaching the stage of enterprise-level availability, while the application of various vertical fields is also in full swing, and towards the final realization of the form of "federation", in the case of ensuring the security, privacy and compliance of data, give full play to the value of data and promote the development of industrial ecology.

In addition, the multi-party computing technology standard on privacy protection is also being further improved. Just this week, the world's first international standard for federal learning was approved by the IEEE Standards Committee, which was led by WeBank to participate in many Chinese companies, and also gave Chinese companies the right to speak in the development of independent and controllable core technologies.