Author | Song Wanxin
On May 17, the 2021 Sohu Technology 5G&AI Summit was grandly held in Beijing.Tian Feng, the founding dean of SenseTime Intelligent Industry Research Institute, delivered a keynote speech as a guest “The Global Change of AI Ethics” . After the speech, he accepted an interview with Sohu Technology. When the rapid development of artificial intelligence and big data, discussions on data security and AI ethics also emerged. Tian Feng gave a detailed answer in the interview.
Tian Feng believes that artificial intelligence is a tool, and a tool must depend on the purpose and the person who uses the tool. Only when users have values. “We hope it will be used by good people, so our so-called AI ethics is more about letting good people use artificial intelligence, rather than letting evil people use artificial intelligence.”
In order to train artificial intelligence, user data is collected more and more, and many cases of privacy infringement have occurred in the process. Tian Feng pointed out that we must not give up food because of choking. “Artificial intelligence will have various risks. If you don’t use it or study its technology, it will never develop. This will bring humanity to industry regression and development. It will always be a theme, whether it is China, the United States, or Europe.” He said.
In fact, while artificial intelligence technology is developing, privacy protection technology is also developing simultaneously. For example, the federated learning technology proposed by Google in 2016 is one of the most popular technologies for privacy protection in the industry. Tian Feng explained the technology and said, “The data is available and not visible. When I judge whether the data is true or false, I only answer a yes or no, and will not tell me the underlying original data. This is the principle of data services.”
Tian Feng introduced that in addition to federal learning, privacy computing, encryption of biometric data, and some corresponding anonymization processing are all security protection technologies that SenseTime Intelligent Industry Research Institute has invested in research and development in various scenarios.
The following is a concise compilation of the interview record:
Sohu Technology: Would you like to briefly introduce to the netizens your research institute of Intelligent Industry and whether SenseTime has some interesting research?
Tian Feng: Artificial intelligence is changing the life and business model of each of us. SenseTime Intelligent Industry Research Institute focuses on the industrial value of artificial intelligence and our social value, including social impact. This is the result of our continuous efforts to write various smart industry reports, digital transformation reports, and global AI policy analysis and ethics-related research results.
Sohu Technology: A common topic, is technology neutral or valued. What do you think of this issue?
Tian Feng: First of all, we believe that artificial intelligence is a tool and a representative of the forefront of technology. The tool must depend on your purpose, so its users, the people who use this tool, must have values.
It has three directions. In general, science and technology, the first is to become a tool of the industry, and the second is to become a scientific, forward-looking theoretical research result. So the second one is the development direction of cutting-edge technology. The third is the unavoidable topic, weapons. During World War II, science and technology developed rapidly. Basically, these three directions evolved. In the end, whoever uses such technology has corresponding values.
Sohu Technology: The technology itself is still a tool?
Tian Feng: Yes, we want it to be used by good people, so our so-called AI ethics is more about letting good people use artificial intelligence, rather than letting evil people use artificial intelligence.
Sohu Technology: User privacy protection has always been a hot topic. In the era of big data, the demand for data is very large. In order to train AI, etc., are there any technologies and rules to better protect user privacy?
Tian Feng: Regardless of China, the United States and Europe, everyone attaches great importance to the privacy of the whole people, which will promote the development of the entire technology industry. The ethical protection of artificial intelligence now has a purpose, first to promote development, and second to protect the bottom line. Artificial intelligence will have various risks. If you don’t use it or study its technology, it will never develop. This will bring industry regression to mankind. The first development will always be a theme, no matter it is China, the United States, Europe.
The second thing is to keep the bottom line. Everyone is like sitting on a boat. If any place leaks, the boat will sink. From the perspective of Shangtang Technology, we have made a lot of investment and work in artificial intelligence ethics. Including the Intelligent Industry Research Institute, our main core task is to do research on the ethics of AI and the joint research and development team to make more technical inputs. For example, what we are talking about now is federated learning, privacy computing, including the encryption of biometric data and the corresponding anonymization processing. These are all kinds of security protections and practices that we continue to do in different industrial scenarios.
Sohu Technology: Federal learning is a relatively hot technology in the industry. Can you explain to the public in simple words what kind of technology is this? Can it protect privacy?
Tian Feng: We often say one sentence, which is particularly suitable for describing this. The data is available or not visible. When I judge whether the data is true or false, I only answer yes or no, and will not tell me the underlying original data. This is data. The principle of service.
Second, federated learning is actually multiple data, and even data from multiple business entities are put together for fusion machine learning training and data cross-analysis. However, this environment can guarantee the security of the underlying data to the data of all parties. From the IT era to the cloud computing era to the artificial intelligence era, there are various technologies to protect the underlying security of multiple data, and more can be obtained from the above. The results of good data analysis, the data of insurance companies and the data of Internet companies are intertwined to produce insurance premium data of thousands of people, which does not mean to publish the private data of every car owner. This is to look at the federal government from an industry perspective. The integration and innovation of the data of multiple commercial entities realized by learning can not only obtain commercial value, but also protect the data resources of each party from being leaked.
Song Wanxin: Shang Tang has invested a lot in this area?
Tian Feng: Yes, we have joined hands with many university laboratories around the world to do artificial intelligence technology and ethics research and development. We have established a computational law and AI ethics research center in cooperation with Shanghai Jiaotong University, relying on the first-line scientific research of universities and enterprises Strength, we will work together to study the technology and soft topics of risk control in various aspects of AI ethics and legal compliance.