On August 30, Mr. Long Wang, founder of Matrix Origin, was invited to share his ideas about data capitalization, trusted computing, and CreDA at the “Elastos Web3 4th” with the Chinese community.
Mr. Wang talk is translated below in its entirety:
Hello everyone, I’m honored to share some of my viewpoints with you online, and I would like to express my gratitude to Elastos, Rong Chen, and Feng Han for their invitation.
Today, I’m going to talk about data capitalization. I’ve communicated with Rong and Feng on topics related to data. We’ve reached a consensus and cooperated in many regards.
First of all, what is the definition of data capitalization? In the real world, we gain knowledge and information from primary school to university and work based on such knowledge, and then receive corresponding rewards. This is a value system in the physical world.
How can the digital world be capitalized? In fact, data is close to people. Taking the increasingly popular metaverse as an example, people’s behaviors in the physical world can be turned into binary data which is then incorporated into the digital world. How does data work in the virtual world? How can we make full use of its value? The metaverse actually exists all the time, but it has no scientific or structured system. For instance, the data about our operations on Weibo or Twitter are sent to a data center for modeling, and the system sends us content and advertisements based on its prediction about what it believes we are interested in.
Nevertheless, there are still differences between the capitalization of the digital world and the capitalization of actual data. Firstly, it is difficult to reliably copy knowledge. We cannot completely copy knowledge, learning, and the ability to apply knowledge, but the data in the digital world can be copied indefinitely, in principle. Secondly, in the physical world, the application of knowledge is highly dependent on people. For example, if you believe that Mr. Chen and Mr. Han can accomplish something, even if they tell you that they have already transferred the knowledge to someone else, you certainly won’t believe that this third individual will be able accomplish the same feat as Mr. Chen and Mr. Han. Your doubts exist because you know that knowledge cannot be reliably copied and transmitted without loss. Besides, there are legal and moral restraints on the application of knowledge in the real world. This phenomenon is explicitly evident in the types of data that can or cannot be used to make money, as well as in the types of data whose use will result in punishment. However, in the virtual world, all operations are completed by computers at a high frequency and on a large scale. Therefore, it is difficult to morally or legally constrain any specific person.
In addition, knowledge in the physical world can be priced basically through work, use, feedback, and other variables. For example, how much does it cost to upload your ID card to the Internet? Unless you are using the black market, it is difficult to define the value. Perhaps only the information of the richest man in the world will be of great value.
So, what should we do? Some people put forward the concept of trusted computing to solve the above problems. How can we accurately manage data according to the behaviors and owners of data and then compute and use the data on the premise of ensuring data security and user privacy? For instance, when it comes to making commercial models, launching advertising, and conducting risk control, data can be effectively used when individual privacy is well protected without being duplicated infinitely. Another function of trusted computing is the capacity to evaluate computed data. During modeling exercises, trusted computing may evaluate the weight of data in the whole model, which will make the model useful, and therefore valuable. Since trusted computing can determine the value of data, it can also support certain trading mechanisms.
Currently, trusted computing can be divided into several schools. For example, in secure, multi-party computation, multiple sets of data can be put together to produce collective results. However, the data still belongs to individuals, and no one is able to use these results to deduce personal information about individuals. Secure multi-party computation currently includes homomorphic encryption, federated learning, and blockchain-based exploration.
Homomorphic encryption is a mathematical approach that functions in much the same way that a Bitcoin private key cannot be deduced from a public key. As a calculation method based on mathematical principles, homomorphic encryption puts together multiple data sets for calculation. It’s impossible for the person who gets the results to deduce any user information, and through signatures, the calculated data can only be used once in each operation. However, homomorphic encryption has its shortcomings, which is to say that abundant encryption means extremely high operating costs – requiring seven, eight or even more than a dozen times more power than would standard calculation operations. As a consequence, although my data is only worth 100 yuan, I may have to pay 1,000 yuan to protect my privacy, which is not practical. Hence, many people are exploring possible solutions. But there is currently no solution that can be feasibly implemented.
Another approach is based on federated learning, which refers to modeling exercises in a distributed system. When optimizing a model, it is the model rather than the original data that is transmitted between nodes. However, there’s a problem: increases in both security risks and network costs arise because calculation operations are not completely reversible processes verified by mathematics. Since specific users can be identified logically, this approach is not a complete instantiation of trusted computing, and its application scenarios are relatively limited.
We’ve been exploring solutions for many years. For example, Mr. Han and I launched the data capitalization campaign to discuss superior forms of implementation. At that time, we had a lot of discussions and discovered a third path: a combination of distributed computing and blockchain. Distributed computing solved the problems of user privacy and data security to some extent, and reduced the cost of trusted computing by consensus computing. What’s more, through this consensus mechanism, blockchain technology’s distributed storage and smart contracts are at least theoretically flexible in terms of cost and protection, thereby achieving a certain degree of balance.
Tamper resistance is another strength of blockchain. Many people attempt to enlarge the value of time-related data. For example, many people try to change their date of birth and medical data, which is of great value to insurance companies. And the longer the time frame, the greater the value of the data. When such data is stored on-chain, the function of blockchains will become more obvious. Tampering with credit data is common in western countries because credit data is very important to companies and individuals in those countries.
Therefore, we cooperated with Elastos’ DID Team to launch the CreDA credit prediction machine project, the essence of which is to explore a path that makes it possible to compute and conduct other operations with data on the premise of protecting user privacy and ensuring data security. The results can provide reference value for various use case scenarios. Our team has done some analyses of on-chain data, and we’re trying to utilize trusted computing to conduct credit evaluation processes based on a proprietary credit model.
Credit scores have entered the blockchain world, and they may be very soon applied to the processes of borrowing and lending. In the future – especially once combined with DID technology – there will be many more application scenarios. Of course, we still have a long way to go. However, since homomorphic encryption, federated learning, and blockchain can coexist, trusted computing will be widely applied in different application scenarios. The future is bright, and we look forward to more in-depth cooperation with Elastos going forward.