Artificial Intelligence has been developed in academia for more than 20 years. Due to the acceleration of computing capacity over the years, it seems that the time when computers will be more akin to human brains is nearly here.

Computing capacity has already passed a threshold. Taiwan is a high-tech economy. Now that AI has seeped out from the labs and is about to turn to industrial applications, we must quickly join the fray. We need to merge our existing strengths with AI. On top of that, we need to let AI be applied across industries.

Without doubt, Taiwan’s strength is semiconductors. In the future, semiconductors will penetrate all fields, and their applications in such areas as healthcare, the environment and intelligent homes will become broader. The more applications there are, the more we will benefit. Emerging industries are another field.

The second important point is that machine learning requires repetitive teaching, just like human learning, which pertains to two points: One is that you need sufficient data to teach it; you need big data. Second is the teaching process. You need deep learning, a move similar to human nerves committing information to memory. Big data and deep learning require computing capacity over a long period of time.

Taiwan needs computing tools, or else we will be much slower than others in AI development. That’s why I have been so vocal in saying that the nation must establish high-capacity mainframes; half of them should be given to academia for research, while the other half should be given to industry for upgrading. As for big data, I will do this by using planned funding for the Internet of Things and open data. I still need to further arrange that part.

Establishing [research] centers is imperative. To cite an example, Taiwan Semiconductor Manufacturing Company (TSMC) is using AI to improve its manufacturing processes. However, it is out of the question for them to spend billions of dollars on the procurement of a super computer, because this is after all not the main thrust of their R&D. All industries in Taiwan, not just TSMC, can rely on AI for industrial upgrading. Therefore, I hope that half of the capacity of these center facilities can be made available for industrial upgrading.

The third part is talent. Aside from mainframes, I want to establish AI research centers to carry out these ideas. If AI is going to be so important in the future, Taiwan must become one of the global AI centers.

In order to become a center, talent and industries must have critical mass.

The goal of establishing centers is that we hope the funding will be enough to enable us to invite world-rank researchers. Since I can hardly change Taiwan’s entire salary structure, I can only rely on enticing people to return by highlighting “emerging technologies.”

Taiwan needs critical mass; we need to entice international talent to stay in Taiwan a little longer and visit more often. Taiwanese researchers can also become part of this pool; only then will we be able to become a location for exchange that is recognized around the world.

I hope to establish three or four large research centers in Taiwan. Taipei, for instance, is a center for digital healthcare and financial technologies. Taichung could make smart machines and assistive devices. The South makes medical instruments and robots.

Using Your Hands

AI does not just appear out of thin air; it must be combined with ICT (information and communication technology). Some AI, like financial technology, is pure software, and Taiwan is definitely weaker there. But Taiwan has always been strong on the ICT front. When it comes to software/hardware integration, Taiwan has a competitive edge. Smart cars, for example, require the transmission of a lot of data. For this, you need software/hardware integration, and that’s an opportunity for Taiwan.

To be frank, Taiwan lags somewhat in AI research. Why have we fallen behind?

AI research managed to overcome hurdles because the people who worked on image recognition for ImageNet [visual database project] hand-annotated more than one million pictures. Therefore, technological research must emphasize a hands-on approach.

People in Europe and the United States work with their hands. When they encounter a problem, they roll up their sleeves and work on a solution. That’s how they find true breakthroughs. We also need to reverse Taiwan’s system for allocating scientific and technological resources.

The field of AI is very young. Stanford University Professor Li Fei-fei, who [in November of 2016] just joined Google Cloud, is just over 40. If these people were in Taiwan, they would probably never even get a chance to say anything. We need to let their voices be heard and their ideas supported.

I have already requested all of the departments in the Ministry of Science and Technology to adjust review mechanisms based on the specific character of the various academic disciplines. We should not only look at academic papers; we also need to check how influential the research is. A lot of new technologies are basically discussed online. No one reads papers published more than two years ago. Our incentives must change, too. We should not look at the papers but at the influence on society. For example, scholars in the humanities could spend time writing specialized books.

I also want to demand that, in the future, project reviews in the different fields of study are no longer carried out by former government officials, university presidents or deans of research and development, because they already control access to resources, so we should not let them control these [research] projects.

Half of the members of the commissions that allocate resources should be younger people under the age of 50. That would be a more dramatic change. (Interviewed and compiled by Yi-shan Chen)

Translated from the Chinese by Susanne Ganz