Gespräch mit Audrey Tang, der Digitalministerin Taiwans


Conversation between Stefan Heumann, Member of the SNV’s Management Board, and Audrey Tang, Digital Minister of Taiwan

Audrey Tang has been the digital minister of the Taiwan Executive Yuan since 2016, which made her the first transgender highlevel official in the executive branch of the Republic of China at the time. Prior to her political career she was one of the youngest coders ever to work in the Silicon Valley, contributing to numerous free software programs, such as SVK and Slash. Stefan Heumann met her in her office in Tawain and talked with her about the equalizing opportunities of decentralized smart grids, the challenges of implementing technological advancements in complex, hierarchical political systems as well as the societal role of AI.

As part of the “Radical Transparency Protocol” the conversation between Audrey Tang and Stefan Heumann was recorded and published. We have edited this transcript in order to improve its readibilty. The spoken word applies.

– Begin of Transcript –

Stefan Heumann: Maybe just some personal background about what brought me here. I’m married to a Taiwanese. We met in the United States, where I got my PhD in political science. So I’ve been coming to Taiwan every year for more than 10 years. She’s a big supporter of equal marriage and other social issues. She has been following those campaigns and has been very disappointed by the outcome of the referendum. From her, and also from other people, I heard about the work that you’re doing.

I thought we also have some overlapping interests in how we work, by trying to reach out to civil society in a collaborative way, to find solutions. It’s exactly our philosophy that we need to find collaborative solutions.

I was really interested in meeting you, hearing from your perspective, learning, and also in asking you, if you have had any engagements with counterparts in Germany and maybe if there are any opportunities to facilitate that.

Implementation of empowering Technology in Rural Areas

For example, what role can technology play in regard to the discussion of energy transition? Germany has done a lot of investing in renewable energy because it is trying to get out of nuclear power. The problem is that, with renewable energy, you have a decentralized system, where you do not control the feed in. When the sun is shining and the wind blows, you have lots of energy, but not necessarily when people want it.

Audrey Tang: Which is why you need a smart grid.

Stefan Heumann: Yeah, and we do not even think you need a smart grid anymore. The smart grid idea from 20 years ago was a centralized idea, that you have some centralized authority that manages the grid. What we really think is that you need a decentralized smart grid.

Give people – we call them prosumers in Germany – directly the ability, who have a solar system or wind system and enable them to directly trade and feed in energy. There is a low transaction cost of digital platforms as a way to implement this. We think that these peer-focused, decentralized systems are really important to solve some social problems.

In the German system we are not there yet on a regulatory level. There are still very old coal industry interests in Germany, and nuclear interests, like you have here in Taiwan, that are pushing back on this, too.

Audrey Tang: In Taiwan, even the pro-nuclear people are also pro-renewable. I think that renewable energy is very much something that everybody agrees on. When we talk about smart grid, there are actually enabling laws and regulations that enable, for example, indigenous communities to completely own the hydro, solar or wind power plants in their indigenous lands.

We’re now figuring out ways for them to trade back with the grid system nearby. I think this is not just about the economy; it is actually a regional revitalization, because they identify with the sovereignty of their indigenous lands.

It’s very important to have it all understood and managed by the local people. It goes beyond the economic argument. It is mostly a social solidarity around the energy that everybody can understand and participate in.

I think both pro- and anti-nuclear people are actually in favor of that idea. It’s actually easier to push now.

Stefan Heumann: The indigenous communities are actually leading this in Taiwan?

Audrey Tang: A lot of indigenous communities are very interested. I have personally visited one, the Taromak community of the Rukai nation. They worship the sun god. Actually, they use the same word for the sun god and the solar panel. It is like the solar panel is an incarnation of the solar spirit.

They have a hydro plant in their region, and they have a few wind turbines as well. Mostly, they are focusing on solar energy at the moment. The three combined together take care of more than 100% of their need. They are now figuring out a way to share the energy with other communities. That will, of course, need a lot of design. Previously, exactly as you said, everything was transmitted to Taipower and they redistributed it. Now, they are working on peer-to-peer relationships. There are some social enterprises formed as local association-controlled companies. There are also some co-ops in this space. There are different structures.

Stefan Heumann: You need to put the policy in place to support them. Is there also some funding that the government gives for this?

Audrey Tang: Yes, of course. The regulations are already there. It’s just the interpretation and operation that are difficult, because nobody has done this before. We need a lot of operational infrastructure to make this happen. For example, there are existing electricity lines, which Taipower has left in that area. Can they reuse those lines instead of setting up new lines? How should the peak and non-peak costs be calculated?

The operation needs to be understood, administered, and governed by the indigenous people in their land, instead of far away, in a control center in a Taipower building. It doesn’t really need regulatory change, but it does need an algorithmic change.

Stefan Heumann: Have you had any interactions with the green movement in Germany on these issues?

Audrey Tang: Sure, yeah, the Green Party of Taiwan is part of the Global Green.

Stefan Heumann: This is not the DPP? The Green Party is separate.

Audrey Tang: That’s right. Maybe I should say "Green as in SDGs" (laughs). There are quite a few concerted movements, as well as shared agenda, with the Global Greens on the local level. The greens, the Green Party and related parties, such as the Tree Party, recently have captured the city councilors’ spots, but not at the legislative level yet.

Most of the political alliances are still on the local level, which I think is good. We’re now seeing a lot of the national development going into what we call the Regional Revitalization of all these different precincts and districts. There are 134 and they make up the majority of the land area in Taiwan, but actually less than one tenth of the population. Many people registered there don’t actually live in these places. A lot of these areas are currently suffering from dwindling and aging populations as well as structures that create a vicious cycle of a lack of employment and a lack of infusion of new people.

Regional Revitalization, which is the main theme this year aims to make sure that the people here get to control their own agenda of local development with the end goal of basically reaching an equilibrium of population within two years. They are starting to have a brain gain instead of brain drain in these areas.

Stefan Heumann: Your role in this is also to organize the engagements with the local population, with this model that you have developed?

Audrey Tang: Yes, exactly. So far, we have used two methodologies. One is the regional tour, where I go to different places and literally have the 12 ministries’ people see through a two-way video conference what the local people’s need are.

Starting this year, we’re also expanding this model to the youth council. There’s a youth advisory council at the administration level. We are also having the youth counselors – many of them are actually local organizers for Regional Revitalization – to invite these people to actually visit their place. Not just across the screen, but actually the locality. The National Development Council is working on the enabling technologies.

For example, many of these areas told us that autonomous vehicles, especially drone delivery, are critical. Otherwise, just getting the necessary material is difficult for them. They have to drive a very long distance, for example. Autonomous delivery is a key. Another one is teleworking. A lot of work here is actually to enable the people to relocate back to their homeland and to work there through teleworking initiatives.

Stefan Heumann: You have fiber in the ground to reach all these communities, right? For telework, you need quite good connections.

Audrey Tang: That’s right, along with 4G LTE coverage. Anywhere in Taiwan, even in the most remote islands like Dongsha, if you do not have 10 megabits per second, it’s our fault. Broadband is a human right. We have been laying out the basic infrastructure for the past couple years. Now, we are reasonably sure the vast majority of these areas in need for revitalization have good bandwidth.

Stefan Heumann: For some remote areas, that is very costly. Is your telecommunications company and that infrastructure private?

Audrey Tang: The Chunghwa Telecom, although a private company, is I think more than 30 percent owned by the administration. They are one of the telecom providers, but they are more willing to absorb the necessary cost, for example, during the 4G band bidding, as well as the 5G in the future. The idea is always that if you get a priority placement, you are expected to also work on the places with less developed, or less profitable to work.

Stefan Heumann: They have a big debate about this in Germany now, because we have an auction on bandwidth that is supposed to be used for 5G technology. Actually, the telecommunication companies are suing the government about the requirements, because they say it is not profitable for them to serve areas that are not very densely-populated, because you wouldn’t have many customers there.

Audrey Tang: The point is that the government is not asking them to operate at a loss. It is asking them to put in the infrastructure so when we do Regional Revitalization, they will actually regain the infrastructure payments, assuming, of course, that some of these regions actually do get revitalized. If you do not couple it with this narrative, of course, it seems as if the state is forcing the telecom to operate at a loss.

It is worth it to put in the upfront investment that enables teleworking in particular. If the government workforce, the public servants working on these projects, all get teleworking from within those precincts and districts they can bring all those data to the locality. Then, ask the local organizers: what is it like to see that region from a government, whole picture kind of view? Then, they can discover their own identity, take the stakeholders, and put on a proposal. Then, very different from the previous government projects, it is on an on-demand, agile workflow. There’s no one-month window for proposing, two-month window for budget, or anything like that. It is basically all real-time service.

Government Efficiency and Political Legitimacy of NGOs

Stefan Heumann: How did you get the ministries here in Taipei to buy into this? Maybe you can explain this puzzle to me. On the one hand, I’m very impressed, because even before you joined the government, Taiwan was already pretty active on open government and open data, and has a very strong record on this.

On the other side, like Germany, you have traditionally pretty hierarchical, formalistic bureaucracies and as you go to the senior level, they tend to be older and more detached from the new technological developments. How do you explain that?

Audrey Tang: I think open government is embraced here because, compared to the NGOs, compared to the social sector, the public sector actually has less legitimacy. I think that is the answer to your question. In Taiwan, we lifted the martial law in the late 80s, but the first presidential election was only in the late 90s.

There is a decade where the regional co-ops, the regional NGOs, and so on, very wellknown names, have been building their legitimacy. Even today, when we go to, for example, disaster recovery or whatever, if the public sector publishes a number and the social sector publishes a number, most people will believe the social sector number. The public sector is working with a legitimacy defict. Because of that, open government is the only way forward.

Stefan Heumann:
 The narrative that I’m often using in Germany is that it is for them. They need to learn. It is for knowledge transfer, too. They should embrace this, because it will make their policies stronger.

Traditionally, they have had their own expertise, and they did everything in-house. To get them to understand that collaboration can actually be better for them, because, when they go outside, they get criticized. They get pushback. They don’t have a very positive image of engaging with civil society.

Not all of them understand that this could be actually positive for them, and actually could make their work more effective. Do you also use this kind of narrative, where it is really about legitimacy?

Audrey Tang: Not much. I usually say, “This is a safe space.” If you go to the Social Innovation Lab to have a conversation, the people, even if they are angry, they are not going to punch you across the screen. (laughter)

I emphasize the safety of the engagement. The safety actually means less risk for everybody involved. That is one. The second is that the credit is shared. Across the screen, in Mandarin, we say 見面三分情 [literal translation: Meet Three Points] — there is 30 percent more trust just by meeting alone. Once you meet somebody in person, it is less likely for the people to start protesting, or with a more ad hominem angle, because they have already met you before. Even across the screen, I think it enables maybe 20 percent of trust. Once people get to know each other, actually, they cannot really be that vicious in their communications.

Stefan Heumann: That is probably also your experience, that you have to bring the people into the same room. The video screen can be a first step. It is important, but really, the real relationship building is for people still to get physically present in the same room.

Audrey Tang: It is two rooms, and each one facilitated by a facilitator. That is what we call connected rooms. In each one, as you said, it is a face-to-face meeting. We’ve been doing this for more than a year now. Overwhelmingly, I think people in Taipei appreciate the fact that they don’t have to go to all those rural areas. It is just me who travels. [laughs]

They also appreciate to learn the actual story. Previously, they just got their stories as two pages of A4 papers, or a few slides of presentation. They can’t really reconstruct in their head, how it is like to be in that locality. They really do appreciate that they are actually all in the same room and listening to the same story.

They do appreciate that, because there’s a lot of back and forth, and real communication. I think this is a more equalizing force between the professionalism of the public service and the local people, who now see them not as anonymous, but actually as a cohort of colleagues who can brainstorm and solve their longstanding problems. Once they do solve the problem, and brainstorm in this atmosphere, there’s a lot of appreciation of the professional capacity of the public servants. Previously, the credit was absorbed by the minister. Now, the credit is given to anyone who actually contributes.

Stefan Heumann: Do you prepare the government officials before you go into the engagement?
do you do a workshop with them to get them ready, explain them, and answer their questions or concerns they have about this kind of format?

Audrey Tang: Very much so. As I said, the risk reducing is really just the entry. Even more important, I think, is to actually prepare everyone so that they learn that these meetings are not there to make them overwork, or to make them commit on things that they’re not prepared to commit on. Rather so, it is mostly to prepare them to shift from a trade-off kind of thinking to a synergy kind of thinking. We do workshops, and in each ministry, there is a team of participation officers whose whole work is to meet with strangers. Every month, we look at what training topics are there in the previous month. We decide on one or two topics to collaborate. Even if a ministry doesn’t have any case, they can still join as co-facilitators, as creators, to the other ministry’s cases, and do capacity building before they actually get a burning issue on their hands.

Stefan Heumann: I think this is all really valuable in terms of also just changing the culture in the ministry and the mindset, as you said, to go from combative to collaborative. What I’ve seen, and what often are the biggest problems of these processes, is how you make them sustainable, and let the people who participate see that there are actually outcomes produced. That there are actually policies implemented, so that this is not just having good conversations.

Audrey Tang: Like we say, the new tax filing system now has 96 percent approval rating. This is a boon to everybody participating.

Political Participation and Transparency through the means of Technology

Stefan Heumann: How have you integrated this in the process that people can track the follow-up from what was discussed at the workshops? How do you do that?

Audrey Tang: We build an accountability trail, an account for the policymaking. The basic idea is that we go back to the Join platform to reply with the tracking updates.

Stefan Heumann: That’s a website where people can check their update?

Audrey Tang: There is the e-participation website. The website actually has more than five million users. With Taiwan being 23 million, it’s actually a quarter of population. The website, very interestingly, is not only used by the administration. Actually, one of the most active users, apart from the cities, counties, and administration, is the auditing department, the Ministry of Auditing in the Corrective Yuan, which is another organ. It’s an independent accountability organ apart from the administration. What they do is that whenever any government, like a city government, tries a new way of doing things, they’re in charge of auditing it.

Previously, if they block that innovation, for example, the Taipei City used to use, for example, subsidy to have employment opportunities for handicapped people. For people with disabilities, there’s a fix set of staff where they can help exactly X number of people with disabilities to work in various duties every month. It’s very easy to audit. You just see whether those vacant positions are filled with people with disabilities.

A few years ago, the Taipei City shifted the experiment with what they call the Social Enterprise Building, the SE Square. It’s a building where they rented, at a normal price – like one Taiwan dollar per month or something – to social enterprises, but they have to recruit people with disabilities. The idea is that the social enterprises will be motivated to recruit more people, because they have more fiscal control and freedom. The people with disability can even join the board as a co-op, or things like that. However, it’s very hard to audit. It’s not very clear, if they have not yet recruited X number of people with a disability, why that is the case. Is it because it’s just ramping up, that it’s dwindling, or how does it even work, over the course of five years?

There’s fear, uncertainty, and doubt in the civil society, as well as in the city council. Now, the Corrective Yuan, of course, is charged in building a new auditing mechanism for this. Previously, if they blocked this innovation, then, of course, they will be charged by the press, saying, “You’re blocking civic innovation.”

If they build an accountability mechanism that’s not really accountable, then of course, the civic councilors will say that they are not doing a very good job to address the concerns. Now, with the joint platform, what they do is that they just say, “OK, we see the city is now trying a new thing. What are your doubts and fears about it?” Usually, we see hundreds of citizens just coming and sharing their fears, uncertainty, and doubt.

Stefan Heumann: Do the people share anonymously?

Audrey Tang: People share pseudonymously. They have to authenticate with an SMS number and an email or a social media login, so we know that they are actually at least SMS-holding.

Stefan Heumann: So that they are really here, resident in Taiwan, and so on? You would be concerned how this could potentially be manipulated from outside, those kind of discussions.

Audrey Tang: Of course. I’m sure that it’s possible for people who really want to troll the forum to get an SMS number from Taiwan.

Stefan Heumann: At least it makes it more difficult for people to do something like that.

Audrey Tang: It’s hard to get 5,000 SIM cards. [laughs] You’d probably get discovered if you did that. At least they won’t flood the forum.

Stefan Heumann: You can get anonymous SIM cards here. This is a big issue in Germany, and privacy is a big concern. As you know, there are – also in Taiwan – discussions about the problem of disinformation, and potentially foreign influence. You have mainland China across the water. If they want to get involved into controversial discussions, that could undermine them.

Audrey Tang: I think with the SMS number, it’s actually kind of difficult to flood the forum. We do get some personal attacks, and so on, but there is a well-known way to fix those, by only responding to the ones that are constructive, and hide the parts that are not constructive. It’s not censorship. You can click to view the previous history if you have too much time on your hands. (laughter)

What I mean is just not to have people’s attention be squandered. We do have pairwise voting, like one column being the pro argument, and another column being the con argument. We have people upvote the arguments that they think are best when it comes to e-petition.

Stefan Heumann: Just for me to get a sense of the scale and the resources, how big is the team working on this and supporting this?

Audrey Tang: My office is, at most, one person from each ministry. At most, I can have 34 colleagues, because we have 34 ministries. At the moment, I have 22.

Stefan Heumann: They wouldn’t build the website and do moderation?

Audrey Tang: There’s a vendor, UDN Digital, building this website and keeping it running. I think it’s a team of 30 or so. In the National Development Council, there is also an operation strategy team that thinks about, for example, what kind of regulatory announcements to make and what kind of budgets to highlight. These are maybe five people. In each ministry, there’s a team of participation officers, as I said. I think on average, it’s three or four people per ministry. Altogether, it’s maybe 200 people, but not full-time. Maybe 100 people or so work full time.

Stefan Heumann: I think it’s important to have your own resources to give out a contract when you need some website development, or some services to implement. Do you work with public-private partnerships, or partner with companies that help you do this?

Audrey Tang: We’re building most of our supporting infrastructure publicly on GitHub. We do get contributions from the social sector, but we’re not relying on them. Just in PDIS, in my office alone, there are easily five people who can code, or five people who can design. If they really want to get something done, I myself am a programmer. We complement the work of the social sector, but we’re not relying on the social sector for this work.

Artificial Intelligence: Challenges and Opportunities

Stefan Heumann: I’m also a member of the German Parliament’s Expert Commission on Artificial Intelligence (AI). In Germany, generally like in all the countries, everybody’s talking about machine learning, artificial intelligence, all the data we have now, what we can do with it. Countries are developing national strategies.

Audrey Tang: 
I gave some comments on the German AI strategy. Der Digitalrat circulated a copy, through what they call the hypothesis system, which is a web annotation document. I know Beth Noveck, the person who brought about this open government initiative thing in the US, is now part of der Digitalrat. I also had an interview with Julia Kloiber, the Founder of Code for Germany.

Stefan Heumann: Yeah, I know her really well. It’s really innovative what she’s doing, and also transforming how our ministries work, who are funding the Prototype Fund.

Audrey Tang: I think she directly inspired the g0v grant, which is again like the Prototype Fund. The g0v people, of course, in traditional g0v fashion, did not accept funding from the government, or from any political parties. This is entirely in the social sector, and I think it’s been running really well.

Stefan Heumann: I think to scale, you have to put public resources behind those things. I think you see a lot of great initiatives in civil society that don’t scale, or are not sustainable, because they run out of money. A lot of foundations only want to give seed funding. Obviously, you know this problem. There are no resources for followup. What is your comment on the German AI strategy, or what is your thinking on it?

Audrey Tang: First of all, I think of AI as "assistive intelligence" and machine learning as collaborative as mentioned in my interview with Julia. I remember pointing out that machine learning is not just an “industry” strategy. It is actually a social strategy as well, if everybody, including schoolchildren, understand how machine learning works, then it is an empowering force of so-called personal computing, because people can feel that everybody can customize and relate to these ideas.

On the other hand, if few people understand it, then even if you build the best privacy by design, you build the best overcoming bias, or whatever into it, just by the lack of literacy, it is still just as good as a top-down machinery. I think the common awareness and education, and especially empowerment through regulatory co-creation, through the freedom to customize, the freedom to self-organize around the machine learning apparatus and to make new norms. I think that really is the key. Education strategy has a lot to say, at least as much as the industry.

Stefan Heumann: The problem in Germany is that in the federal system education policy is made on the state level. The national government cannot mandate certain education policies. Some states are better than others in terms of trying to integrate new approaches and new knowledge into the curriculum, and others are really not doing that.

The national AI strategy, however, was very much driven by economic and industrial concerns. The first chapters are all about that, how you can have the best research, and how you can transmit that into industry and startups. That’s the core.

We actually need to start it. With the process I described to you, we actually tried to start that conversation around an AI strategy in our organizations. We were also thinking about how to democratize AI, and give as many people as possible access to it. This is obviously a question of the accessibility of the technology. Is it all going to be owned by corporate entities, or do we continue at the moment? A lot of this is open source available.

Audrey Tang: One of the open source examples is the one of self-driving vehicles. All of them are open hardware and open source. When the local people want to accommodate their flow – for example, when they shop at the flower market and self-driving tricycles follow you and form a fleet. Basically, adjust their distance to people, and signal their internal state through a way that people can understand. Like there are people with feedback saying, “This is almost like a cyclops. It’s not very friendly.” The latest iteration – because it’s open hardware, you can just tinker it – now uses two eyes. They have a cheap LIDAR and optical sensors, which is why they can now make eye contact. (laughter)

I think that’s a very good idea of people seeing self-driving vehicles, and think of these as things, instead of truck fleets.

Stefan Heumann: Taiwan also has an AI strategy. How would you describe how they address the social apsect? What bothers me about the discussion in Germany is that they always talk about the Chinese AI strategy, but they always only talk about what’s done in Beijing. Nobody looks at innovative approaches towards AI that are coming out of Taiwan.

Audrey Tang: Again, we put a lot of emphasis on education and regulatory co-creation. We allow the social innovators through Regional Revitalization planning and touring, to basically declare their own hometown’s need of one particular intelligent technology, like drone delivery. Then, they get to break existing laws and regulations.

Stefan Heumann: They can get some exeptions from national regulations? They will need to get the approval from the national government to do such a test bed, right?

Audrey Tang: That’s right. Basically, you need to have an application in what we call the sandbox system, which is not just for AI, of course. It’s also for fintech, for platform economy. It is a shared front end for anyone who wants to break the law for a year. You go to here and file in your lawbreaking proposal. (laughter)

Don’t think of it as lawbreaking. It’s like forking, having your own version of the law, and test it for a year. Everybody can apply, saying whether it is a good idea or not. By the end of the year, people may want to extend it to test in another setting.

Stefan Heumann: The main applicant would be a city, a local community, a city or a council or can it also be a private company?

Audrey Tang: It could be a private company, and even a person with its own business plan. We don’t discriminate between the organizations. Of course, the more multistakeholder you are, the better. The end result is to get the multistakeholders to decide collectively whether this experiment works for the local revitalization or not.

Stefan Heumann: How do you determine whether or not it works?

Audrey Tang: A consultation process. Usually, we use things like which can highlight the people’s divisive arguments, as well as the consensus statements. It is itself AI-powered. Anyone can go in and see their place among the different clusters of their neighbors. Then, they can propose their feelings for everybody to vote. People can vote agree or disagree. As they do, their position moves toward the people who feel likewise.

Stefan Heumann: This is very interesting. This basically means that you need a lot of people familiar with this moderation and collaborative process, right? Have you created some sort of academy, where you train people in these kinds of competences?

Audrey Tang: On the local level. Actually, the District Office, is going to be the coordinating agency. Of course, there are many district offices. In our public service academy, we have an e-academy. We recorded our collaborative workshops and our training materials. I personally trained around a hundred, no, actually 900 or so people, even before joining the cabinet. Of course, after joining the cabinet, we trained even more people. We recorded it and put it on the e-academy. This methodology spreads by itself, because each collaborative meeting, which lasts for five to six hours, is by itself a training program.

Stefan Heumann: That’s also my experience. You cannot just tell people theoretically, “This is how it works.” They need to do it.

Audrey Tang: It’s training by doing, that’s right. Basically, this is what we call osmosis. The more you get into this participatory system, the more you can feel confident when it’s your turn to run it. That’s the basic idea. If the stakeholders end up having a consensus around this particular technology, then we’re committed to merge it back into the regulation, so that every other region can use it as well. If the experiment ends up with people’s consensus saying, “This is not actually a good fit,” again, nothing is lost.

We all learn something, and the new innovators can try something else. It is almost like a monopoly, actually. If the legislation decides that this needs a legislative change, they can take three or four years to actually pass a law that remedies the shortcoming they see from this experiment. During those three or four years, the business model itself is still legit. The experiment extends to the point where the legislation is happy with it. You have a de facto monopoly for three years.

Stefan Heumann: What’s also great is that you make the law based on real experience and testing. This is one of the problems: We make a law, and then we look, “Oh, what happens?” Then we see, “Oh, it doesn’t work out like we wanted it to be.”

Audrey Tang: Policymakers cannot regulate something we don’t have first-hand experience with. Unless we all ride the self-driving tricycle, how are going to suppose to make useful laws to govern them? Our AI strategy is regulatory co-creation. That is one of the five strategies here. Talent and democratization are others.

Stefan Heumann: Is education another one?

Audrey Tang: AI Pilot is similar to DARPA. At the moment, it’s managed by the Ministry of Science and Technology and the Board of Science and Technology. It’s not defense, [laughs] but it is science and technology. At the moment, they’re running a grant challenge.

Stefan Heumann: We are discussing this in Germany, too. There probably will be a similar agency too, although we are not very comfortable with the military angle in Germany. We will create a civilian DARPA in Germany.

Audrey Tang: Finally, we work with the MSME (Micro, Small and Medium Enterprises) for innovative approaches. The industries need innovation. Various upcoming industries in Taiwan, many of them MSMEs, face challenges that could be automated.

For example, the water company, the Water Corporate of Taiwan, said that the water leakage detection is one of the most time-consuming parts of their work. Their experts, using a listening device, have to circulate Taiwan for a year and a half before new leakage is discovered. They really need machine learning to help them to identify the ones that are most likely. They use these as training material for AI academy and other talent training. It’s not just solving puzzles for practice, but actually to solve real challenges, as posed by the MSMEs.

What’s important is that this is not asking for a perfect solution. Even if your solution only improves the efficiency by five percent or so, it is still a real gain for a MSME. It also makes the talents much more likely to integrate back to the industry, instead of just going elsewhere. I think this is a pretty good solution to do it AI-driven by essentially training experienced designers that can redesign their workflows for full integration.

We already have experience working with people with mental neurodiversity. If we’re going to integrate people with Down Syndrome or people with handicaps into the workflow, you have to redesign the workflow so that they can work on the part that they are good at, without burdening them with the part that they are not good at.

Now, we are seeing AI as a kind of people, who need to be integrated into the workforce. It is the same design of service work that needs to be implemented. We integrate those kinds of designs into the MSMEs. Allow them to outsource the parts that are trivial, that really need automation, and then introduce the AI students as the solution givers to partner with the MSMEs. That’s the other part of the five pillars.

Stefan Heumann: Wow. You’re really ahead of us. I knew it before, but now, talking to you, I can really see that. It’s very impressive. Also, how the things are connected. I really like that, how the thinking is very integrated. That’s really cool.

I’m glad you’ve heard about the German Digitalrat. One of the challenges of the Digitalrat is that it’s not permanent. They are coming in from the outside. What we really need is somebody like you, who really is in day-to-day contact with the ministries, and driving innovation. Hopefully, the Digitalrat can get our government to adopt those kinds of approaches, to make it permanent, and scalable across government. The challenge has not been that there has been a lack of innovation. We’ve been talking about open data, open government, and things like that for a while in Germany. Our challenge is implementation. For implementation, you need to talk about collaboration. You need to talk about mind shift, culture shift, working differently, then changing things, how you used to work. What would you personally describe as the biggest challenges?

A lot of people in the German government will tell me that the biggest challenges for them is, we have 50 years of bureaucracy, so there is lots of rules that determine how things need to be done. It’s very difficult for them to step outside of these rules and formal processes that have been designed over many decades, and do things differently.

Audrey Tang: We don’t think it’s a large problem here, mostly because even though we’re also a continental law jurisdiction, we explicitly carve out parts that are good for experiment. The sandbox laws are essentially the legislation saying, “It’s OK to break the law for a while.”


Stefan Heumann: That’s very hard for Germans to adopt that kind of thinking.

Audrey Tang: That’s right, but I think it’s really worth it. If you don’t have a clear-cut sandbox, then of course, there are literally hundreds of obscure rules and interpretations that can block your way. You don’t discover them until you run into them. I think a canvas, or a sandbox, with strictly set geographic and/or time boundaries, I think that is really the main innovation vehicle that we’re working on. 5G for instance: everybody understands that 5G is good for something, but they are not quite sure for what. At the moment, we are encouraging dozens, if not hundreds, of different vertical experiments on 5G, but always test open spectrum that’s good for maybe a couple years. Then if they found that 5G is really not the problem, we really need to have some fiber instead, then maybe it’s not a very good fit for 5G technology.

Stefan Heumann: Do you have open spectrum?

Audrey Tang: That’s right.

Stefan Heumann: We’ve been pushing for that in Germany, too, and it’s very, very difficult to get.

Audrey Tang: The tests need to be limited by geography.

Stefan Heumann: I know. There are interference problems. Because there are lots of spectrums opening up with the switch to digital TV and things and we are auctioning everything off.

Audrey Tang: To the telecoms?

Stefan Heumann: Yeah, and the telecoms are very keen to get it, of course, but we have said there should be some spaces for experimentation. Look at Wi-Fi. We wouldn’t have that innovation, range of devices, and use cases, if we didn’t have an open spectrum that everybody could use.

Audrey Tang: At the end of the day, it’s good for telecoms as well. The experiments may fail, but at the end of the day, the telecoms are best equipped to scale out those innovations. They have to go back and talk to telecoms if they want to expand it everywhere.

04. Januar 2019