Shulou(Shulou.com)06/02 Report--

2019-09-01 12:23:52

Dr. Shen Xiangyang, Global Executive Vice President of Microsoft and head of Microsoft artificial Intelligence and Microsoft Research

At Microsoft Research (MSR), we are always trying to solve problems about the future-we strive to advance the highest level of computer science, we quickly transform and integrate the latest technologies into Microsoft products and services, and we are also helping customers around the world to incubate disruptive technologies and create new business models and solutions.

Microsoft Research is located all over the world, with branches in Redmond, New York, Cambridge, Massachusetts, Montreal, Canada, Cambridge, England, Bangalore, India, and Beijing and Shanghai, China. In the past year, Microsoft Research around the world has made many great achievements and made great contributions to the development of Microsoft itself and even to the progress of the whole field of computer science. For example, we developed the first method to store data coding in DNA, which can compress a data center the size of a warehouse to only a few dice; our Project Natick project built a data center at the bottom of the ocean; and we developed an open source solution that can explain machine learning to help people open the black box of artificial intelligence. We have also created a tool called Code Jumper that uses modules of different shapes to help visually impaired children learn how to program.

When we founded Microsoft Research in 1991, we were determined to look to the future, constantly searching, and creating major breakthroughs that could affect the future. To achieve this goal, we have developed a strategic roadmap spanning 10 to 20 years from today to the future: we design and prepare products and services that may be on the market within 18 months; we have clear directions and requirements for projects in the next 3 to 5 years; and most importantly, we will pay more attention to technology trends in the next 10 to 20 years.

The development of HoloLens holographic glasses is the most representative: more than 20 years ago, we began to invest in the basic research behind HoloLens; with the emergence of Kinect sensors and KinectFusion, we achieved a real breakthrough, using 3D sensors to completely reconstruct the 3D structure; and then HoloDesk, which can assemble 3D models into a virtual 3D environment. Over the past three or four years, the technology and application of HoloLens have matured, and a major breakthrough in mixed reality has been achieved-today, we demonstrate how to use HoloLens in conjunction with neural network text-to-speech technology to create a holographic stand-in that can speak multiple foreign languages-exactly the same as in science fiction movies.

Being prepared in advance represents Microsoft Research's thinking about the future. Today, a series of technological innovations represented by artificial intelligence are making many incredible things come true. However, we should not only promote technological progress, but also seriously consider the possible social impact and the potential challenges that mankind will face together, prepare in advance, take precautions, and build responsible artificial intelligence from scratch.

Dr. Shen Xiangyang, Global Executive Vice President of Microsoft and head of Microsoft artificial Intelligence and Microsoft Research, presented at the symposium on "Computing in the 21 Century".

In the technology industry, Microsoft is the first to put forward the determination to build responsible artificial intelligence. In 2016, Microsoft CEO Satyanadella published a column on the shared responsibility of artificial intelligence. a few months later, Microsoft's principles of artificial intelligence were publicly presented for the first time: fairness, reliability and security, privacy and security, inclusiveness, transparency, and accountability.

We have published the book Computing the Future, which makes a comprehensive and detailed discussion on the moral, legal thinking and social impact that may be caused by the development of artificial intelligence. Microsoft has established the AETHER (abbreviation for "Ethics of artificial Intelligence in Engineering and Scientific Research") artificial Intelligence Ethics Committee, which is composed of Microsoft leaders, engineers, researchers, lawyers, etc., responsible for discussing and evaluating issues related to artificial intelligence ethics within Microsoft. Today, Microsoft has always adhered to six artificial intelligence principles in the process of developing and applying artificial intelligence, and every product we launch has to go through the ethical review of artificial intelligence.

We have made some progress in eliminating AI bias through technological innovation and ensuring the fair and inclusive development of artificial intelligence. For example, our research team has solved the problem of gender bias in text search using a natural language processing tool called word embedding.

The so-called word embedding is an algorithm used to convert words into vector digits. it can assign a corresponding vector number to each word based on the massive text data from news data or web page data. By comparing the correlation between common words and the gender pronouns "he" and "she" in the vector coordinate system, the researchers found some obvious features, such as "sassy" and "knitting" are closer to women, while "hero" and "genius" are closer to men. The reason why the algorithm gives gender characteristics to these words is that the benchmark data set used by the training algorithm-- usually data from news and web pages-- itself, has a "gender bias" caused by language habits. the algorithm also naturally "inherits" gender differences in human understanding of these words.

As a result, when we use the experimental algorithm to speculate about her career based on Melinda Gates' LinkedIn information, we get the judgment of a "teacher", but as long as we replace the personal pronoun with "he", Mrs. Gates's profession will become a "lawyer". Today, many AI systems use this word embedding mechanism. In order to solve the problem of gender bias caused by this, we propose a simple and easy solution: in the word embedding, delete the judgment dimension that distinguishes "he" from "she". For more details on this scheme, we have published a paper, and the relevant data are also shared on GitHub for everyone to use.

Scientific research only respects the future, so as scientists, our mission is to challenge reality, explore new knowledge, and create differences-- which is the key to the success and development of Microsoft Research for decades. At Microsoft Research, we believe that the so-called future is created by our constant exploration. What we need to do is to guide, innovate, enlighten, and create a better world for all mankind by promoting the transformation from today to tomorrow.

Https://www.toutiao.com/a6731553062860620291/

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.