Data protection and distributed AI

Data protection and distributed AI

Staff Writer
Staff Writer
Pinduoduo Content Team
October 20, 2020

Our technology infrastructure adopts a distributed artificial intelligence (“Distributed AI”) approach, which we believe in the long run would significantly improve data security on our platform and afford better privacy protection to our users.

With the rapid growth of online shopping, the amount of data generated and exchanged on e-commerce platforms is increasing dramatically day by day. Meanwhile, users are increasingly concerned with the security and privacy of the information they supply to the platforms. Issues such as the ownership and right to use these data have become more contested.

Most of today’s e-commerce platforms use a centralized artificial intelligence approach to store and manage data generated on their platforms. The predictive modelling, user behavior simulation and advanced data analytics on these platforms are conducted on a centralized cloud with data extracted and uploaded from each user’s smart device. All the data that is constantly generated by users are stored in the centralized cloud systems, which are generally available to platforms with little restrictions, posing the risks of information leakage and making them prime targets for cybersecurity attacks.

At Pinduoduo, we believe that users should have full control of their own data and should be able to actively direct the application of their data to better serve them. Users should be able to decide what information can be uploaded onto the cloud and what should be kept privately on the local devices.

As computing power at the edge improves and becomes more powerful in the mobile internet era, more AI functions can be hosted locally on smart devices. More data can be stored and analyzed by our smartphones, without the need for centralized processing. The AI on local devices act as personal agents for users to handle most of the data analysis. Only with users’ specific authorization will certain data be uploaded onto larger platforms where developers could have access and use them to refine algorithms and improve these “local agents” to better serve the users.

These local agents could integrate the relevant private information and public information to produce recommendations to their users on their local devices based on its analysis of the users’ behavior and preferences. They could also communicate with each other to exchange information for optimizing the recommendation system without transmitting private data to the centralized cloud.

As such, sensitive data could be stored securely in local devices, and algorithms could be developed and refined in the centralized cloud system based solely on data voluntarily disclosed by users to the public without data breach issues. The bottom line is that users should have control and oversight over what data can be shared with and used by whom, when and how.

Based on these beliefs, we have set up our technology infrastructure under a Distributed AI approach. While it is still in its infancy, we see great potential for its future development as it could optimize our user experience by providing them more comfort and sense of safety when they shop on this online marketplace.

POSTED IN:
POSTED IN: