At the World Artificial Intelligence Conference, which ended not long ago, the top 20 list of the five circuits of the 2019 World Artificial Intelligence Innovation Competition (AIWIN) was finally announced.With the “intelligent interactive products – overdue elves”, Tongdun Technology has become a list of companies in the “AI + Finance” track.Tongdun Technology is a 36-inch continuous reporting company. Founded in 2013, it is currently positioned as an intelligent risk control and analysis decision-making service provider. It will provide risk control, anti-fraud and analysis based on artificial intelligence + big data analysis services.Decision service.The company’s team includes members from companies such as Ali, PayPal, UnionPay, FICO, SAS, Ping An, McKinsey, etc. Today, the company has over 1200 employees and 80% of its members are product development and data scientists.Tongling’s business idea is “Intelligent Analysis as a Service”, which currently serves multiple industries and scenarios including finance, internet, logistics, big health, retail, and smart cities.At present, the number of customers with the same shield service has exceeded 10,000, of which more than 5,000 credit customers, the daily API call volume exceeds 100 million, the daily new data volume is nearly 10 billion, and the stock data is 35PB.In April this year, Tongdun Technology officially announced a new round of over $100 million in financing.In March of this year, Li Xiaolin, a tenured professor at the University of Florida, served as the vice president of Tongdun Technology and the dean of the Institute of Artificial Intelligence.Continue to strengthen the capabilities of AI technology in machine learning platforms, federal learning, knowledge mapping, intelligent voice and computer vision.At the World Artificial Intelligence Conference, 36氪 also interviewed Li Xiaolin, who introduced us to new developments in AI technology.The following is the content of the interview, edited by 36 editors: Q1: In the past year, what new products did Tongdun Technology launch?What aspects of AI technology research are being promoted?In 2018, Tongdun launched heavyweight products such as the enterprise-class service platform “Qi Yu”, the intelligent voice platform “Hertz”, and the machine learning platform “Tianji”.Our artificial intelligence research institute and the strategic direction of the same shield are closely related. The current research focus on the artificial intelligence technology is mainly focused on deep learning, including federal learning, reinforcement learning, intelligent speech, computer vision, natural language processing.several aspects.Q2: The same award is the intelligent voice robot. In the direction of intelligent voice, what are the main grounds in the same shield?Our intelligent voice robots are also made for the financial track.We provide cloud services, customers can customize the appropriate robots on the cloud platform according to their needs, you can customize, you go to lead a conversation, ah then you can go with the robot, think of something, want to doThis kind of thing, then we open this platform, then everyone can match it, in this way, the product can also reach very precise customers.The intelligent voice robot can complete the work of traditional credits such as credit business due diligence, key business information acquisition, credit smart approval, risk identification, service return visit, etc. In the collection field, it can provide intelligent over-delivery service, and can complete the collection without human input.Telephone; in the insurance field, it will provide insurance companies with voiceprint core, as well as service return visits, claims customer service and other services.It should be emphasized that the intelligent voice robot with the same shield is strictly controlled in terms of compliance.(Editor’s note: The AIWIN selected for the top 20 products “Overdue Wizard” is one of the products of the intelligent voice platform “Hertz”. It utilizes speech recognition, natural language processing, speech synthesis and other technologies to collect various standardization services after loan collection.Intelligent transformation, the interaction process is completely transparent, supervisable, and traceable.) Q3: You said that federal learning is the ongoing research and development focus of the same shield. What kind of technology is federal learning?Federal learning is a kind of distributed machine learning/deep learning technology. The participating parties build a common virtual model on the basis of encryption (can be the same or different), and the data of all parties in the whole process of training and interaction always stay.Locally, does not participate in exchanges and mergers.Federal learning is decentralized algorithmic logic, so no one party can have all the data, no one has all the models, share open data, and not enjoy the data, which can maximize data security and data privacy.Q4: Why focus on federal studies?The original intention of federal learning is to solve the problem of “data islands” and also to solve the problem of privacy compliance.From the financial industry, every bank, insurance, securities and other financial institutions have their own databases, accumulating a huge amount of data.But the data is on their own servers, and the data is invaluable, but because of the competitive partnership and data compliance issues, the data is difficult to get through.Formed a “data island”, not only in the financial industry, but also in logistics, retail, medical, government and other industries.Tongdun Technology first proposed the concept of “cross-industry joint prevention and control”, so that the data connectivity of different enterprises in different industries is just needed.And Tongban is an independent third-party service provider, and there is no competition with all parties, so it is also very suitable for doing this.Federal learning does not require communication between raw data, thus ensuring data security and privacy.Imagine if the models of risk control of various financial institutions can be interoperable and the data can be opened. Can it prevent more kinds of risks?Q5: How to prevent the system risk of federal learning?We can use the means of encryption and blockchain to prevent the problem of parameter leakage. Even if the attacker’s technology is extremely clever, the parameter leakage will be very limited.At the same time, we are further developing more secure algorithms and mechanisms that can resist malicious attacks. Q6: At present, has the federal learning of the same shield been applied? What new development will be in the future? The federal learning will be the same.The strategic focus of Shield Technology R&D has been explored in the fields of smart credit, micro-finance risk control, anti-fraud, marketing, etc. It has already reached cooperation with several major banks and is exploring transnational federal learning methods.In the future, we will also open an open AI operating system. We will provide our AI technology and AI products as an operator, and we will also introduce third-party AI service products, which can be understood as an open “application mall” for AI services.This is also in line with the trend of “open bank”. Related reading: 36氪 starting | “Tongdun Technology” completed a new round of over US$100 million financing, dedicated to exclusive interviews on risk control and intelligent analysis decisions |Unicorn, “Tongdun Technology” completed the $72.8 million C round of financing to complete the B+ round of “Tongdun Technology”, began to arrange acquisitions to help enterprises solve network windsRisk and fraud, Tongdun Technology received nearly 10 million US dollars of A+ round financing, broadband capital led investment to help companies solve network risks and fraud problems, Tongli Technology, founded by former Ali and PayPal anti-fraud experts, announced that they have obtained IDG and Huachuang.Capital 10 million yuan investment image source: Tongdun Technology official website.