Hans Uszkoreit : Interpreting the Future of Artif i cial Intelligence

By Jessica Xu

我的中國(guó)故事MY CHINA STORY

Hans Uszkoreit : Interpreting the Future of Artif i cial Intelligence

By Jessica Xu

Photo by Ni Tianyong

International Talent Monthly: In March 2017, you and your team visited Beijing E-Town Industrial Zone. It is reported that you are planning to establish an Artif i cial Intelligence (AI) Institute. Could you introduce the current progress of the project to the readers?

Hans Uszkoreit: We are still in the very beginning stage of this project. The goal is to build an AI technology innovation center that can speed up the transfer of research results and innovative software products into manufacturing and service industries. The center will contain several components: research, avant-garde product development, venture capital investment and technology consulting.

We can have a quick start because we are not beginning at level zero. We decided to concentrate for the time being on a variety of technologies developed in Germany and elsewhere in order to have an immediate bridge from the research lab into fruitful applications. We have started hiring experts but we have also decided to first survey the most pressing demands in the local industrial scene before deciding on our priorities with respect to areas and specif i c technologies. We have already talked intensively to more than 30 enterprises and are getting now a better feeling for the existing challenges and opportunities.

International Talent Monthly: As Director of German Research Center for Artif i cial Intelligence, what cooperation have you had with China?

Hans Uszkoreit: Since 2004 we have had a joint Chinese-German research lab at Shanghai Jiao Tong University.Between 2006 and 2008 I also coordinated a Sino-German project on the topic of digital Olympics, which resulted in a spin-off and a number of software products. Over the years, I have had many bright Chinese students in my German undergraduate and graduate programs including several Chinese doctoral students who completed excellent PhD theses. I also had several outstanding Chinese researchers in my labs. It turns out that mixed teams of Chinese and German experts can produce amazing results. They worked better than teams with only German or Chinese researchers.

International Talent Monthly: As you know, China started rather late in the fi eld of AI. In your opinion, what are the advantages and disadvantages that China has in the development of AI?

Hans Uszkoreit: Yes, China had not been very active in AI for decades but is now catching up extremely fast and successfully. However, there are only few experienced intellectual senior leaders with a broader vision.

The bright young researchers focus on the latest methods that are in high fashion in order to get attention and for their papers to get accepted for international conferences and journals. Older and existing methods that are in high demand and very avant-garde and original lines of thinking do not yet get the attention they deserve. This is a problem for the entire fi eld but it is even more critical in China.

The Chinese society has strong focus on high quality education,which is paying off now. There are large numbers of young ambitious candidates for our fi eld who have a solid basic education in math and computer science. The evolution of a strong and broad AI scene is slowed down by a lack of senior experiences in teaching and research leadership and by the immense pressure due to the academic evaluation system. Quantity of publication counts more than quality of training.

On the other hand, China is a great place for technological innovation because of the scope of the market and the economic ambition. Avant-garde products often need initial markets of early adopters before they can fully mature into larger markets. By the sheer size of the Chinese economy, these early-adopter markets are much bigger than that in Europe and even North America, which allow for faster adoption here in China. Chinese are also fond of new technologies and are willing to try new things. This is true both for consumers and technology managers in enterprises.

There is another advantage that China has. Most of the world's internet markets are dominated by a few US-based enterprises. In China many of their business models were adapted to the Chinese market by domestic corporations. This parallelism slowed down development for a while but turned out be benef i cial for China in the end, because in this way China developed and owns much highly valuable competences that in the Western world are all concentrated in a few US fi rms.

International Talent Monthly: You are expert in the fi eld of natural language, but it is believed that natural language understanding is a relatively marginal discipline decades ago, why did you choose it as your research area?

Hans Uszkoreit: Indeed, Natural Language Processing used to be a small and marginal area in science and technology but today it has become a central fi eld of research. Language technology is bridging the gap between humans and machines. Soon, all kinds of technology will be easily controllable by voice interfaces. Language technology also bridges the language boundaries between nations and cultures since machine translation keeps improving. And finally natural language processing is going to enable effortless access to all knowledge in the world. Search engines are becoming knowledge machines. Searching for documents will be replaced by fi nding the right answers.

Therefore Google, Baidu, Bing and Sogou are mainly language technology enterprises.

But actually, I am not only working on human language but also on big data, combining and analyzing textual, numerical and visual data. And we are working on large repositories of machine-usable knowledge that has been extracted from texts.

International Talent Monthly: What do you think of your research experience in Stanford University Laboratory?

紅蜘蛛是一種喜歡群居的病蟲，并且具有假死性，因此利用此病蟲害的特點(diǎn)，可采用震落加灌溉的方式淹死部分蜘蛛；同時(shí)還需要重視田間的雜草清除工作，尤其需要重視清除禾本科雜草，以此使病蟲蟲源得到相應(yīng)減少。還需要做好藥物防治，常用藥物有、阿維菌素、掃滿凈等。5000-6000倍液的1.8%阿維菌素以及每畝3-4g的5%甲維鹽可溶性顆粒劑等，均可取得較為良好的紅蜘蛛蟲害防治效果。

Hans Uszkoreit: Studying and working in the US had a huge influence on my career. I got to know some of the greatest scientists, inventors and entrepreneurs in IT and I learned a lot from them. Although Stanford and Silicon Valley is in the United States, it is a truly international zone and it is closer to the future we have in mind than the existing society in the US.

International Talent Monthly: In your research career, what impressed you most?

Hans Uszkoreit: I am most impressed by the fact that many technology fantasies that we know from science fi ction movies are now becoming reality. Self-driving cars, robots that walk and talk, and access to the entire knowledge of theworld from any place on earth......and soon in any language. And I am proud that my research associates and I were able to contribute to such developments.

International Talent Monthly: In 1997, a world chess champion was beaten by IBM's Deep Blue, even causing the panic of human beings. On May 23, 2017, AlphaGO from Google won against Player Ke Jie in Wuzhen Town, China. What do you think of this historic event? Will it indicate a new direction for AI?

Hans Uszkoreit: These spectacular events are impressive but the events themselves cannot change the world. However, they clearly signal a profound change. Now machines can master some intellectual challenges that are too hard for 99.99% of people, such as chess and knowledge quizzes. AI can also perform better than the average humans on many other tasks, such as automatic translation, face recognition and driving a car. So, for all of these tasks, AI exhibits superhuman performance. But for many tasks that are much simpler, AI still fails. And even the best AI systems cannot combine the tasks. The translation system cannot drive, AlphaGO cannot translate and the translation system cannot answer a single question about the translated texts. But people can do hundreds of different tasks.

Nevertheless, AI is now ready for hundreds of applications.

Hans Uszkoreit: By considering, weighing and exploiting huge amounts of information in very short time, AI can help people to make better decisions. It can help investors choose among potential investment targets. It will help doctors fi nd the right diagnosis and choose the best therapy. It can help the product manager improve products by getting a better understanding on the needs of consumers. Business executives, engineers, politicians, technologists, journalists and all other professionals who have to digest and consider large volumes of information, will greatly benef i t.

Since more and more big decisions are made by groups today, such as committees and political bodies, we are now also developing AI-supported mechanisms for better collective decision making. Such programs also accumulate relevant information but they also help in sharing, clarifying and collecting arguments and factual evidence.

But AI is also going to help people to make better decisions in daily life, such as asking oneself which route do I take to work today? Which product to buy, which doctor to consult, which retirement plan to select.

But will we trust AI to make decisions for us in the future? This depends on the risks involved. When it comes to routing decisions in autonomous driving or when to replace worn-out parts of a machine, we will trust technology and we are already doing this today. For instance, many decisions of banks concerning the acceptance of a credit card payment are already done by AI software. But when it comes to serious medical, economic or political decisions, I do not see.

International Talent Monthly: With the development of AI, more and more jobs will be replaced by machines in the future. Is that a good or bad sign for humankind? As ordinary people, what could we do in order to survive in the competition against AI?

Hans Uszkoreit: This is clearly an important concern. AI will not only affect physical labor but also intellectual jobs below a certain level of responsibility and creative demands.

There is a little dangerous in the near future, though. On the one hand, automation and robots will continue to replace human labor , business software and other AI-enriched IT. But countries that have a large number of robots in daily operation usually have very little unemployment. IT and especially AI will not only abolish jobs but also create many new ones. However, these jobs require much higher skills than the disappearing ones. Social welfare system and careful management of human resources need to reduce temporary hardships. The educational system needs to prepare the younger generation for the new types of jobs.

But I deeply believe that in the longer future, eventually all or nearly all jobs will disappear that involve repetitive or uninteresting work.

Employees who dream of quitting work after winning in the lottery or people longing for retirement long before they reach the end of their professional life, usually work on jobs that will not survive the AI revolution. Society still has some time to prepare for such a world, not very much time though. In industrial societies with a good welfare system, millions of unemployed people have the means to live a simple yet decent life but most of them are still not happy. On the other hand, these societies have a serious shortage of people who care for the elderly, help in preschool education, assist career couples with child care and household chores. Numerous old people die in social isolation. Today’s market economies will not be able to accomplish the needed redistribution of labor without some careful social reengineering. AI could play a strong role in matching demand and supply of human labor when the appropriate policies are in place.