Li Xiaolin, tongdun Technology: hope to build Bell lab in AI field

category:Internet
 Li Xiaolin, tongdun Technology: hope to build Bell lab in AI field


After returning to China, there was a lot of choice, but Li Xiaolin felt that tongdun technology and his research in recent years were very relevant, and the application scenarios were also very matched, so he joined tongdun.

Li Xiaolin said to Netease technology that his current key responsibility is the companys Artificial Intelligence Research Institute, and his main work is to explore the basic theoretical algorithm of machine learning and deep learning, conduct in-depth research on the core technologies of artificial intelligence such as computer vision, speech language processing, federal learning, and push the application of intelligence to the market, striving to build it into the field of artificial intelligenceu201c Bell Laboratories .

At present, tongdun technology has a key project, National Science and technology innovation 2030 new generation artificial intelligence major project team, which will participate in the project of generalizable domain knowledge learning and computing engine. Netease technology expresses its expectation and concern for this project.

Blockchain is a hot topic recently. According to Li Xiaolin, tongdun technologys investment and innovation in blockchain field have made great efforts. In fact, as early as 2017, tongdun technology started the pre research and exploration of blockchain technology in the risk control scenario. The two technologies selected by the state Internet information office to publish the blockchain record list are based on the same shield technology blockchain platform IASC. A set of special network infrastructure management mechanism is implemented, which is used for security sharing and collaboration on representative identity records. Through this dedicated network, commercial banks, tongdun and fintech enterprises are in the loan, payment and credit markets. Sharing information safely in an unprecedented way greatly reduces the cost of credit transmission and reduces the numerous verification links of information authenticity.

However, the efficiency problems caused by decentralization have been perplexing the development of blockchain technology. The lab has developed a special identity management blockchain network, using the hyperledgerfabric blockchain development framework, relying on the micro block architecture and queue system, it has overcome the main limitations of current blockchain technology, data processing speed and throughput level, far beyond the traditional public blockchain, in addition, it can safely handle peak and uneven requests.

With the rapid development of science and technology, people are most worried about the security issue. Li Xiaolin said that the security issue is a key link in the current industry development, and the data island and privacy protection issues are the prominent problems encountered in the large-scale industrial application of artificial intelligence.

Data and information security has become an important cornerstone of trust in the digital and intelligent era. Federated learning is a theoretical technology of multi-party data fusion and machine learning model integration through decentralized collaboration under the premise of user data privacy protection.

Through the in-depth exploration of federal learning technology, they will take advantage of the federal idea to bring financial enterprises, such as banks, insurance companies, credit companies, consumer finance companies, operators and Internet companies into the federal framework, defuse the problem of joint modeling among multiple subjects, multiple data sources and different feature dimensions, so that all parties of enterprises can protect each others data privacy Under the premise of cross platform and regional co creation value.

Whether its artificial intelligence, 5g, blockchain and other fields, the talent problem is the most serious and important problem. To this end, Li Xiaolin said: technology can bring real visible changes, and talent is indispensable. In the future, tongdun hopes to introduce more high-quality talents with background in deep learning, machine learning and so on. At the same time, it also needs more programming experts.

In addition, Li Xiaolin also said that tongdun science and Technology Association will regularly give training lectures to colleges and universities, and the school will send some interns to study and work, so as to make some contributions to the cultivation of high-tech talents.

Every young person has different interests and preferences, so the choice direction will be different. If overseas talents want to go back to China to start their own business, its better to join a start-up company like tongdun, first feel the speed and welfare of China, and then start their own business. Li Xiaolin also said that tongdun technology is also actively building artificial intelligence laboratories in the United States and Singapore to attract outstanding talents from around the world.

Their biggest wish is that everyone can realize their self-worth and let their own technology create more social value.

The following are the contents of Netease technology interview with Li Xiaolin:

Netease Technology: we know that President Li has been engaged in AI research abroad. As an internationally renowned scholar in the field of artificial intelligence, can you introduce your previous research direction and achievements?

Li Xiaolin: I graduated from Zhejiang University with a masters degree from Professor Feng Peien, a famous design expert. At that time, I studied intelligent CAD and expert system, which was the era of ai1.0. Later, he received his PhD in information and communication engineering from National University of Singapore and his PhD in computer engineering from Rutgers University of the United States.

The research team I led at that time was one of the earliest research teams in the world that integrated cloud computing, big data, large-scale deep learning and Internet of things research.

Our R & D team has led the development of a number of software platforms based on artificial intelligence technology, among which deep cloud and cognitiveengine are one of the earliest large-scale parallel deep learning platforms in academia. Deepbipolar is the first application that combines deep learning technology with schizophrenia detection. Primatei AI was used to detect the pathogenicity of gene mutation, far exceeding the current best result of 76% to 88%, which was published in the Journal of natural genes. Foldingzero developed by us is earlier than alphafold of deepmind to break through the independent deep reinforcement learning from scratch.

More than 150 peer-reviewed conferences and magazine papers have been published, and 5 international patents have been obtained. Many research results play a positive role in promoting the development of the whole technology industry. Led the founding of nsfcenter for big learning (CBL), the first national deep learning center in the United States. The main research includes deep cloud intelligent platform, deep eyes computer vision and deep reinforcement learning, deep business natural language understanding and intelligent assistant, deep health intelligent medical, deep security system, Cyberspace Security and other projects.

Netease Technology: what kind of opportunity makes you want to return to China for development? Why did you choose tongdun technology?

Li Xiaolin: more than 20 years ago, during the period of artificial intelligence 1.0, when I was studying in Hangzhou, I found that the domestic research resources were limited, so I decided to go abroad to master the most advanced research. In recent years, the field of artificial intelligence in China has developed rapidly, and there are more landing scenarios, so it is decided to return to China for development.

Things that should be tossed around in the United States have been tossed around. I think it should be changed. I decided to go back to China. I didnt know the specific business here when I first contacted with tongdun. After a while, I found that it was very related to what I had done in recent years. The application scenarios were also very matched, so I joined tongdun.

Netease Technology: what do you think is the biggest psychological change when you enter the business world from the academic world? What do you have in common? Whats the difference?

Li Xiaolin: in my opinion, mentors in the United States are like CEOs of start-up companies. They are in charge of more than a dozen students, do invention projects, write articles, and seek support in foreign cooperation. Sometimes it is very difficult. For a certain research, we have to explain the purpose and plan of our team to each other after several setbacks. When I arrived at the company, it was the same. Recently, I ran many companies to persuade them to understand how important our products are, just as when I was doing project research, I persuaded those companies to give support. The academic community is more forward-end, most of the time regardless of risk, just to see if new ideas can spark. But in the company, we need to make a little bit of landing, and we must make a successful product. More often, the business community should be linked with laws and regulations, economy and market, and consider the actual situation more complicated.

Netease Technology: which business are you in charge of at present in tongdun technology? What projects are currently under development?

Li Xiaolin: at present, I am mainly responsible for the companys Artificial Intelligence Research Institute, exploring the basic theoretical algorithms of machine learning and deep learning, conducting in-depth research on the core technologies of artificial intelligence such as computer vision, speech language processing, federal learning, etc., and pushing the intelligent application to the market, striving to become a Bell laboratory in the field of artificial intelligence.

In addition, the company has several laboratories jointly established with the outside world. For example, the artificial intelligence Joint Laboratory of Zhejiang University, the intelligent voice laboratory of West University of technology, and the joint innovation laboratory established with CCB and CCB, etc. At present, one of the key projects is National Science and technology innovation 2030 new generation artificial intelligence major project team. We will participate in the generalizable domain knowledge learning and computing engine project.

Netease Technology: previously, founder Jiang Tao also said that tongdun will continue to increase investment in talents and technology reserve in intelligent voice & voiceprint, image, NLP, machine learning, deep learning and other fields. What can be shared at present?

Li Xiaolin: Yes, Dr. Wu Jianxiong, a top voice expert in the world, is also invited as the vice president this year. Before joining tongdun technology, Dr. Wu Jianxiong served as the vice president of group research in nuance, a famous American company. Nuance is the originator of global intelligent interactive technology services. Under his leadership, he has worked for many famous global enterprises, such as apple, Samsung, Mercedes Benz, BMW, Audi Bank of America and others supported the development of intelligent voice assistant.

In addition, our research institute also has a large number of domestic and foreign top talents in AI field, including intelligent voice experts, computer vision experts, natural language processing experts, in-depth learning and federal learning algorithm experts, with a masters degree and a doctoral degree.

The research institute has established four core R & D groups, namely, intelligent voice, natural language processing, computer vision and in-depth learning, and has made a number of patent achievements, deeply participated in the formulation of industry standards initiated by national ministries, central banks, etc. In addition to basic theoretical exploration and core technology research, the Institute has made gratifying progress in promoting the application of intelligent technology. Based on tongduns in-depth understanding of the financial services industry, the research institute has developed a number of intelligent dialogue robots for financial services scenarios, which can effectively communicate with different interlocutors by adopting different dialogue strategies with highly anthropomorphic pronunciation.

Netease Technology: with the increasing attention from all walks of life to the protection of personal privacy, data and information security has become an important cornerstone of trust in the digital and intelligent era. I want to ask President Li, what do you think about data security? How does your team or tongdun technology solve the security problem?

Li Xiaolin: we believe that security is a key link in the development of the industry. Data island and privacy protection are the prominent problems encountered in the large-scale industrial application of artificial intelligence. While federated learning can effectively protect data security and user privacy, it can also break the limitations of data island.

Data and information security has become an important cornerstone of trust in the digital and intelligent era. The emergence of Federated learning technology will solve the contradiction between data privacy and data sharing. Federated learning is a theoretical technology of multi-party data fusion and machine learning model integration through decentralized collaboration under the premise of user data privacy protection.

Through the in-depth exploration of federal learning technology, tongdun will integrate financial enterprises, operators and Internet companies such as banks, insurance companies, credit companies, consumer finance companies, etc. into the federal framework by using the federal idea, defuse the problem of joint modeling among multiple subjects, multiple data sources and different characteristic dimensions, so that all parties of enterprises can protect each others data privacy before To realize cross platform and regional co creation value.

Netease Technology: when it comes to data security, we have to mention blockchain technology, which is also a hot technical topic recently. We know that recently, two technical schemes of tongdun technology have been filed in blockchain by the state Internet information office. Can you talk about tongdun technologys investment and innovation in blockchain?

Li Xiaolin: as early as 2017, tongdun technology started the pre research and exploration of blockchain technology in the risk control scenario. The two technologies selected in the blockchain filing list released by the state Internet Information Office are based on the blockchain platform IASC of tongdun technology.

Blockchain is a data storage unit that builds a distributed database system on point-to-point network and encrypts it with asymmetric encryption algorithm. All digital assets, digitized assets, or digitalized assets can be circulated, transferred and traded on the Internet using blockchain technology.

Based on this technical feature, the first test project of tongdun intelligent risk control North America laboratory is to provide users with a dedicated digital identity network mechanism using blockchain technology. This network helps to promote the development of user business, improve the fraud management process, help users track customers in all channels, and adapt to the changing behavior mode of customers.

At the same time, it also implements a set of dedicated network infrastructure management mechanism, which is used for security sharing and cooperation on representative identity records. Through this special network, commercial banks, tongdun and financial technology enterprises can share information safely in the loan, payment and credit markets in an unprecedented way, which greatly reduces the cost of credit transmission and reduces the numerous verification links of information authenticity. In addition, the efficiency problem caused by decentralization has been plaguing the development of blockchain technology. The lab has developed a special identity management blockchain network, using the hyperledgerfabric blockchain development framework, relying on the micro block architecture and queue system, it has overcome the main limitations of current blockchain technology, data processing speed and throughput level, far beyond the traditional public blockchain, in addition, it can safely handle peak and uneven requests.

IASC solves the separation of data ownership and use right by using asymmetric three-stage layered cryptography technology. At the same time, it supports the transparent calculation of data in a safe and isolated sandbox environment, providing a sandbox environment that is ready to use and ready to use. In the sandbox, it solves the invisible and available security computing environment of data, and provides effective technical and security support for the security application of data in various industries.

Netease Technology: from the fields of artificial intelligence, 5g, blockchain and so on, the problem of talents is a relatively concerned problem at present. What do you think of it?

Li Xiaolin: technology can bring real visible and tangible changes without talents. In the future, tongdun hopes to introduce more high-quality talents with background in deep learning, machine learning and other aspects, as well as more programming experts. Everyone can realize their own value and create more social value with their own technology. We have cooperation with Zhejiang University, northwest Polytechnic University, Hong Kong University of science and technology and other domestic universities. We often communicate with each other. We regularly go to these universities for training lectures, and the university will send some interns to study and work. I hope to make some contributions to the cultivation of high-tech talents. Young people have different interests and preferences, so they will choose different directions. If overseas talents want to go back to China to start their own business, its better to join a start-up company like tongdun, first feel the speed and welfare of China, and then start their own business. We are also actively building artificial intelligence laboratories in the United States and Singapore to attract global talents. Source: editor in charge of Netease technology report: Zhang Zutao, nt5054

Li Xiaolin: technology can bring real visible and tangible changes without talents. In the future, tongdun hopes to introduce more high-quality talents with background in deep learning, machine learning and other aspects, as well as more programming experts. Everyone can realize their own value and create more social value with their own technology. We have cooperation with Zhejiang University, northwest Polytechnic University, Hong Kong University of science and technology and other domestic universities. We often communicate with each other. We regularly go to these universities for training lectures, and the university will send some interns to study and work. I hope to make some contributions to the cultivation of high-tech talents.

Young people have different interests and preferences, so they will choose different directions. If overseas talents want to go back to China to start their own business, its better to join a start-up company like tongdun, first feel the speed and welfare of China, and then start their own business. We are also actively building artificial intelligence laboratories in the United States and Singapore to attract global talents.