Skip to content
Abstract: [NetEase intelligence July 28th news] Professor Yang Qiang, published a speech entitled the challenges facing AI and the opportunities brought by the transfer of learning, and talked about the plight of the big data and the solutions to the development of AI.
Yang Qiang said, we are in the era of big data driven AI, people will digitize their work, and then improve their efficiency through automation of AI. However, now we have more and more encountered the problem of data, many enterprise data are isolated island can not get through, at the same time restricted by the EU GDPR, such as the constraints, increasing this dilemma.
Professor Yang Qiang specifically explained the two plight faced by big data.
One of the difficulties is privacy, security and supervision.
Taking the GDPR of the European Union as an example, GDPR explicitly stipulates the right to be forgotten of the user and prohibits the use of automatic model decision making. It has a great impact on machine learning, because it is not easy to let users agree to use it, and the use of independent decision is to satisfy three points, including the necessity of the contract and other laws. The data body clearly agrees.
Under the general trend of data privacy regulation, it is not easy to solve this problem. Professor Yang Qiang put forward the idea of federal transfer learning. This idea hopes to establish the enterprise ecology of machine learning. Each enterprise has no local data, the effect of the model is constant, and a virtual model is established in the case of no violation. Professor Yang Qiang said that using federated transfer learning encryption technology, collaborative modeling, learning model process does not exchange user numbers, does not infringe on privacy. Another big dilemma is the plight of small data. Professor Yang Qiang proposed transfer learning, and cited examples of cross domain public opinion analysis. Yang Qiang finally said that faced with the plaguing of data development, we hope to use federated transfer learning technology to establish joint modeling solutions to overcome data barriers. On the basis of legal norms, a consensus mechanism is understood by all participants to ensure safety compliance. For example, in the financial field, the federal transfer learning alliance of financial industry can be established. (Xiao Yi) pays attention to the NetEase intelligent public address (smartman163), explaining the big events, new ideas and new applications in the AI field.
Under the general trend of data privacy regulation, it is not easy to solve this problem. Professor Yang Qiang put forward the idea of federal transfer learning. This idea hopes to establish the enterprise ecology of machine learning. Each enterprise has no local data, the effect of the model is constant, and a virtual model is established in the case of no violation. Professor Yang Qiang said that using federated transfer learning encryption technology, collaborative modeling, learning model process does not exchange user numbers, does not infringe on privacy.
Another big dilemma is the plight of small data.
Professor Yang Qiang proposed transfer learning, and cited examples of cross domain public opinion analysis.
Yang Qiang finally said that faced with the plaguing of data development, we hope to use federated transfer learning technology to establish joint modeling solutions to overcome data barriers. On the basis of legal norms, a consensus mechanism is understood by all participants to ensure safety compliance. For example, in the financial field, the federal transfer learning alliance of financial industry can be established. (Yi)
Pay attention to NetEase intelligent public address (smartman163), to interpret the big events, new ideas and new applications of large companies in AI field.
Source: NetEase intelligent responsible ，editor: Pan Qingqing _NBJS5830.