-
Let’s get started. How can I help you?
-
As I wrote, artificial intelligence is now on everybody’s mind, right?
-
How is Taiwan doing in terms of research and application of artificial intelligence. Of course, US, China, everybody thinks about those countries.
-
Of course.
-
Then a long time, nothing. Then Japan, Korea. Taiwan is on nobody’s monitor.
-
Mind, right, except on the multinationals, who are all very insistent on setting up AI centers here.
-
I know there’s some things going on, but what’s the vision on Taiwan? Where do you see Taiwan right now in the whole setup in the world, and what’s the vision for the next 5 or 10 years?
-
The vision is really simple. It’s innovation, collaboration, inspiration. We see AI as something that’s in the flow of the current Taiwan’s 5+2 industrial focus. You would know that the first of the 5+2 is AIoT, or AI powered Internet of Things.
-
We call it Asia.SiliconValley, because in Asia, trade is very complementary nowadays. The intermediate goods just move around countries. Anything, including this iPad, is made in the world. Taiwan specialized.
-
As you mentioned, semiconductor is one, but also in the smart machinery as well as in the creative use of the IoT devices. The IoT devices is not just deployed in the industrial setting, but also in the citizen science setting.
-
Taiwan, I think...It’s unique in Asia, where there’s a inexpensive air quality measurement box, air box, then you automatically see thousands of people just connecting together their individual measurement devices, because they care about their air quality and they, frankly speaking, want to trust their neighbors’ numbers more than the government’s numbers. [laughs]
-
Very quickly you can see this kind of open data platforms arising by people donating data into the commons essentially. This configuration is really unique, and it also means that whenever there’s a new emergent algorithm coming from the AI research, it is not only channeled to the private sector but also the social sector as well.
-
The social sector takes care of creating their own data, curating their data, making the data legitimate through distributed ledgers and so on, and co creating the norms around the use of such data. I already showed you two of the prime examples of mobile AI stations. These are what we’re saying when we mean regulatory co creation.
-
It means that we collectively set the societal expectation of what to automate, when to automate, where to automate, and then shuffle our curriculum, our AI training program, such as the AI Academy that’s going to join us shortly to basically educate people in responsible use of AI, where they learn about AI.
-
They learn it in the context of solving a real social problem using co created data across sectors, so that everyone can learn to be a data steward in a data collaborative. I think that is our main vision, it is essentially democratizing this vision for AI based technologies.
-
That’s the very high level overview, but that’s the guiding principle.
-
In Japan, they talk about Society 5.0.
-
To integrate, to include the elderly, especially. The same in Taiwan, of course we’re not as old as Japan, but we’re getting there. It’s the same direction.
-
Taiwan kind of has a natural niche in the batteries, in the optics, in the IoT as I mentioned, that forms part of this AI mobility platform so whomever wins at the end as the brunt of AI mobility including actually drones. I went to Tokyo for the Uber Elevate conference where they want to make those vertical take off and landing devices that essentially takes people from the top from one skyscraper to another skyscraper.
-
Again, the battery power, the many of the chips and so on, they are part of the Taiwan. They could only be sourced in Taiwan, because it’s a rapid iteration. People really want the chip design, the interaction design, the field study, ethnography, experience design and so on to make in very closely knit cycle, because nobody really know what’s the right formula, you just have to swarm and experiment a lot.
-
Taiwan is kind of...We have something in each of those areas. People really, I think, Google bought HTC’s cell phone department, but then turn it into the AI research lab it is, because they really want to tap into this integrated fast iteration prototyping process. Whatever brunt it has in the end, I think Taiwan always really have a niche in the kind of being white label part of this creative ecosystem. That’s what we already have.
-
I think what we also want to sell, which is another very new thing, is that we’re a very good sandbox for AI regulatory co creation. Anyone can break a law or regulation for a year, if they think their AI application is good for public benefit.
-
Case in point, there’s a what we call telecom operator that partnered with a bank that says using AI, we can give the younger people who never have traded with the bank in any relationships, we can figure out how much to loan them within their risk limit, simply by looking at their mobile telecom payments history.
-
They don’t really need to go over the counter for KYC, because when they get their SIM card, we already check their identification. In essence, they can just start banking through their mobile apps in the sense of inclusive financing, because they introduce a new algorithm to calculate the risk factors in loaning to those students.
-
We say, "Yeah, but have..." Before we turn it into regulatory text, "Have anyone really your code?" They are like, "We can try for a year on 5,000 people. How about that?" Then we calculate the risk and say, "OK, go ahead."
-
If there’s impersonalization or any people gaming the system, it will be detected early on. The data will be shared. It will be open innovation. If it works, then we’ve merged the norm back to regulation. If it doesn’t, because it’s open innovation, somebody else learns from it.
-
This kind of sandbox is systemically breaking existing laws and regulations. I think it’s something that Taiwan has to offer. For AI based mobility, we also have the Taiwan CAR Lab in Shalun, a green energy city, that can serve as a proving ground and simulation.
-
Not only on the technology, but on the social configurations, their reactions around AI mobility technologies, just like the self driving tricycles and their relationship with the lab here.
-
No, it’s open. It’s generally available now. The CAR lab is generally available. If you want to make a trip to Shalun, you can already see the self driving vehicles just roaming around in the simulation field.
-
Here, it’s the Taiwan C Lab, and Shalun in the Taiwan CAR Lab, C A R, connected, autonomous, and road testing. It looks something like this, because it’s literally just outside a high speed rail station.
-
This is not just for technical testing. It’s actually for social acceptance testing. It’s one of the more exciting things in working policymaking, is just to look at how the society want, in terms of regulation. Instead of saying, "Oh, it would work," or, "It wouldn’t work," just try it out and see it actually working, to some degree.
-
Then collaborating on continuous integration, lawmaking.
-
At what stage will you put it into real use?
-
Usually, after a year of sandbox testing. It can be extended to up to two years, but then we need to make a collective decision as a society whether this autonomous vehicle is good or not. We also use AI based conversation to make such a judgment.
-
Once such a judgment is done, then we determine whether we incorporate those learnings into a regulation, or whether it’s really not a good idea, and it should be tried in some other way. It’s not limited to cars.
-
It could be cars that fly, ships, or whatever. Just autonomous vehicles.
-
What about data protection? You were just talking about this student film thing?
-
Yes.
-
You have to have access to all this data.
-
In the aggregate, of course. Not the raw.
-
The fintech AI, so it’s already used?
-
Yeah, it’s live.
-
Yeah, we have an EU style privacy protection act, our PIPA. We are in the process of getting GDPR adequacy. It should be real soon now. That means that we only use such data, either voluntarily, opting in, or in the aggregate, without any way to identify.
-
The Taiwan Bio Bank, yes, Taiwan Bio Bank. It’s in early stage of partnering with a non profit called Taiwan AI Labs. AI Labs is nonprofit, and it’s founded by, I think, director of Cortana, Microsoft speech AI, called Ethan Tu, who also happens to be the founder of PTT, Taiwan’s Reddit, basically.
-
A very popular social sector, hosted by National Taiwan University, I would call it bulletin board system, a BBS. Crowdsourcing and respect for privacy, while making collective intelligence work, is always the main ideas in AI Labs.
-
AI Labs has three focus, healthcare, smart city, and human interaction. It’s all on their website. I don’t need to go into details. The main idea is that being a non profit indeed, what we call a social enterprise they were able to offer very competitive hiring conditions.
-
People trust the AI Labs team to preserve humanity with privacy spirit, while in the same sense, working with cancer research or other bio bank research. It is a, I think, ethics first approach, really is very helpful in developing such kind of cross sectoral trust.
-
Basically, building relationships. Japan and Taiwan is similar in viewing data not just as material but rather as beginning of a relationship. Actually, GDPR is moving to this way of thinking as well.
-
What would you say is Taiwan’s unique contribution? If you compare it with US or China?It’s all data driven. Taiwan is small islands and not so many data available.
-
I wouldn’t say so. I mean, the citizen science, as I mentioned, is a proof that if we distribute and democratize the data stewardship then we’re not relying purely on a centralized data collection which could be expensive. That is why data is likened to oil in the first place. It’s expensive to collect and extract. Otherwise, it has no other similarity with oil. [laughs]
-
It’s a bad analogy except that it’s expensive to extract and to collect. If we democratize the collection and if we use deep learning and newer techniques for analysis, then it’s neither expensive to collect nor to analyze, to extract.
-
Essentially, by building artificial intelligence on top of our existing plan on industrial innovation based on collective intelligence, we’re saying that an AI doesn’t need to be a centralized power. That is Taiwan’s unique contribution in that our main employment is provided by miss meets by small and medium enterprises.
-
Each of them has a very different configuration of social problem they need to solve in order to go about any business. It’s less vertical integration. Rather it’s more swarm like behavior to solve emerging issues both domestic and abroad. By empowering them to utilize AI and automate whenever they could appropriately, we get the AI talent from all sectors instead of only by CS majors.
-
We get AI talents by essentially the miss meets looking to reinvent their flow without relying on trickle down from the largest corporations as in other nearby countries. I’m not saying we’re just better. I’m just saying it’s a different form of innovation.
-
What we’re saying is that we’re essentially a partner. Whatever your centralized collection device or application is, there’s some part...It could be the chips. It could be the acoustic or optics. It could be the edge computing devices. It could be anything that is invariably designed or produced in Taiwan.
-
What we’re saying is that we’re not particular saying we’re locking in in particular data monopolies as you talk about. We’re not particularly betting any surveillance capitalist regimes. [laughs] Rather, whatever they’re using, first we’re in the loop because they use Taiwanese components.
-
Also, we make sure that our own regulation is done in a way that provides ample opportunity for newcomers, for miss mets to build on the existing base and to, essentially, do social innovation. I’m not blindly saying, "Social innovation is great. You can disrupt the large monopolies."
-
What we’re saying is that we’re happy to work with large monopolies. They’re all setting up AI centers in Taiwan. Once our talent understand the logic in which they operate, then they take those logic and also apply it to solve real social issues.
-
That’s why there’s a lot of start up initiatives in Taiwan going on.
-
That’s right.
-
Tell me about talent. That’s probably now the main headache of all companies or countries that they don’t have enough talent. This AI fever is pretty huge. I know in Taiwan there’s a lot of very clever people and very mathematical and natural science oriented.
-
It used to be that Japan is the strictest when it comes to immigration for white collar talents. Taiwan is not as strict but still pretty strict. Singapore used to be the most liberal with the Gold Card Visa and everything, entrepreneurship programs, and so on. In Tsai Ing wen’s government, we basically copied everything Singapore has to offer.
-
We wrote our own Gold Card 4 in 1 visa. Singapore is three year. We’re three year but renewable. We also made sure that we have a special foreign talent law that makes it very clear that if you are one of the needed, being brain drained talent areas, including art, you can work in Taiwan without having to first find a employer.
-
For people with this kind of talent, we’re essentially location independent. We can work anywhere. It doesn’t matter for us. What we’re saying is, for those digital nomads, when they come to Taiwan, they don’t have to have a employer.
-
They can still work for their own LLC somewhere else. They just enjoy the food here. [laughs] Gradually, over the course of three or five years, they can build a deeper cultural relationship with the people here.
-
Also, maybe after five years or so, would consider to be also Taiwanese. We’re making it really easy to go through it. Actually, you were around in Taiwan in the ’90s. In Taiwan, we have a national ID number. For foreign people, it’s the residence certificate. They look different. The number is different. The second digit is a digit for nationals but the English alphabet for foreign people.
-
It makes it very obvious that foreign people are foreign people even if they have permanent residence certificates. We’re fixing that. Starting next year, the second digit will be normalized. It will look the same format for both people with residence certificate and the nationals.
-
It’s A1 or A2 something, for nationals. It will be A8 or A9 something for foreign people. The national EID will look very similar. It’s all touch NFCs and PKI, just like Estonia, for both domestic and also foreign nationals. That will create a much more welcoming system for people to try out Taiwan for a while and to be also Taiwanese after a while. We’re getting pretty good feedback on those programs.
-
Brain drain to, especially, Silicon Valley but also other innovation centers has been reversed in the past couple years. Ethan Du came back from Microsoft. He didn’t come back alone. He brought his friends, and teams, and so on with him.
-
We’re seeing now that Facebook just today opened its new research headquarter here in Taiwan, the Facebook Taiwan HQ. Of course, Microsoft convinced 200 or something AI researchers here and so on. All of the fame is eyeing Taiwan to be their AI talent.
-
What about German companies?
-
That’s really I think [laughs] something we should improve on.
-
Yes, and we’re happy to share, but we also like, after our period of sharing, that they come back.
-
We’re surely open to competition, but we also have one of the better trade secret and copyright protection laws around. People would usually have kind of a non competing clause. This is all part of modern world.
-
Taiwan is not better, not worse, in terms of this kind of protection, but I think really, what convinced people to come to Taiwan for essentially entrepreneurship, after a while in working with large combative companies, myself included, is really the meaning making, the place making aspect of it, the social aspect of it.
-
I think people are looking less about just having the cutting edge innovation and [Taiwanese] it for a local counterpart, because we don’t have a great firewall to foster that kind of innovation. The PRC is really good in doing that.
-
We don’t have that, right? What we do is kind of a catalyst between different data paradigms. The GDPR paradigm, the US minus California paradigm, and of course, the PRC paradigm, are very different in how the data concentration of power is distributed, managed, and so on.
-
Taiwan is maybe only like Hong Kong, and unlike anyone else, regularly, symbiotically, deal with the three data paradigms. I think the meaning making part in taking an innovation and merging it with another data paradigm, and make some new meanings out of it.
-
That is just plain impossible to do in its original habitat and vice versa. I think that is one of Taiwan’s main attractions if you are an AI entrepreneur.
-
What’s the catch?
-
The catch is that people have heard of Thailand, but not Taiwan. [laughs] That’s the main catch. We have a very good story. We could really do better in telling that story.
-
Would you say that now those AI entrepreneurs or digital people, they go more to Taiwan because it’s notexpensive to live?
-
Yes. The cost of living is one thing. As I said, a digital nomad can live anywhere. Mostly, it’s only time zone diametrics. All places in the same time zone is the same for us. What makes Taiwan the preferred destination in GMT+8, plus or minus one, is, as I said, the social interactions.
-
Cost of living, of course, is one. The food is one. Broadband as human right is easily overlooked but also explains why a lot of people go to maybe Yilan and Taitung and so on. Traditional, if you were a surfer, you would prefer those places. It’s excellent view and so on. Now, thanks to broadband as human right, you can also be here and enjoy 100 megabits per second without any interruption.
-
It becomes preferred place for people to not just host retreats but actually have entire labs there. Because of the broadband, it suddenly solves the discoverability problem. People can still discover you even if you’re literally in the mountains or by the sea. We don’t see that in many of the more upcoming Asian countries. They haven’t solved the broadband as human right there.
-
You already put this in the constitution that broadband’s a human right?
-
Mm hmm. It’s Tsai Ing wen’s presidential platform. We put it in what we call the forward looking infrastructure plan. Anywhere in Taiwan, actually even outside Taiwan like in Dongsha Island or something, if you don’t have 10 megabits per second, it’s my fault. Being held accountable like this is very important as well.
-
This is even without 5G.
-
That’s right. It’s with 4G and cable technologies.
-
When is your target of 5G being available?
-
You mean generally available? It’s next year like everybody.
-
Technically.
-
Technically, of course, there’s already sandboxes just like any other country where setting, again, one year if you’re for profit, renewable if you’re not for profit then make testing fields already.
-
You can go to selected verticals Foxconn may be one of them to enjoy 5G in those limited areas and test whether one of the profiles in the current R16 fits your demand. General availability for telecoms and so on, that’s next year.
-
In Taiwan, we benefit from the geography. In the municipalities, there’s sufficient amount of people. We can still afford competition even the auction price. In the rural areas, people generally agree to share their infrastructure anyway. The geography defines how the broadband is naturally distributed. We still have a pretty healthy ecosystem of two or three plus two or three operators.
-
Do you think with 5G there will be a huge step up to have even more difference?
-
5G certainly makes the number of simultaneously connected devices exponentially more. That’s not the main attraction. The main attraction is on very selected use cases. I usually think of two use cases. One is the worker operatings that construction machines to take down buildings and so on currently with 4G because it’s outdoors. It’s impossible for them to do it safely from a control center.
-
They have to be in that place and with all the workplace safety issues to operate such heavy machinery. With 5G, it becomes then possible for them to incarnate themselves in those dumb terminals and essentially remote pilot all those dangerous workplace operations.
-
When the construction workers do their piloting that way, then it creates enormous possibility for AI as a co pilot to see what they see, to hear what they hear, to understand their decisions, and remind them of their safety issues that may have been overlooked.
-
Essentially, train apprentices, AIs as pilots instead of having to wait for something 100 percent perfect to replace humans. The co learning thing is one of the missions that 5G provides. That is one part. The other part, 5G is necessary is the low latency use of autonomous driving.
-
Currently, autonomous driving is either low speed or in dedicated lanes that can already provide the 5G like connectivity. For massive amount of cars, random placed out of nowhere, 5G is still needed for the latency of mobility. Everything else, we can fiddle with WiFi and fiber and make it happen.
-
A huge amount of money is put into start ups. In TSIthey told me this is very well financed.
-
Yes. The start up is extremely well financed in Taiwan.
-
Which is a very good thing because this is a development everywhere.
-
Yes. Taiwan has traditionally been all miss met when it comes to job anyway. A very large percentage, 70 percent or something, provided by miss mets. Miss mets partner naturally start ups. Large verticals, less naturally. The configuration of our miss met is predetermining a friendly to startup culture.
-
If you put 100 in, how much would come out?
-
Yes, we do have some idea. The best answer is that they approve it. We encourage them to fail fast. One of the start up entrepreneurship classes in universities is just three months. The student has to start a company and do some business and close the company. Hence, put their finance record as the company closes. [laughs]
-
Basically, we want to get this idea of fail fast and provide post mortems and contribution to the ecosystem. When you fail, you fail in front of everybody. Everybody learns something because of your failure instead of always having to be acquired or bought. We also encourage acquisitions but only when it makes sense, not randomly.
-
Taiwan’s main attraction in startup scene is not particularly the ease of getting loans or the ease of getting your credit even before you release your product to being assured by the miss met fund or whatever. We have kind of pioneers, some of those designs, the e tree and things like that. Now pretty much everybody else has those designs as well.
-
It’s still mostly the culture of not afraid to fail. That is the main strength. That can be compared more easily with Japan, which is still finding the normativity in their curriculum for a parent to not panic when their child decide to become a entrepreneur. They have a huge challenge to culturally solve for parents to acquiesce their children to be entrepreneurs.
-
You talked about the sandbox example. Taiwan, as well as Japan, are not very much known for being un bureaucratic.
-
Yes. That’s the new story we need to tell. We design a lot of innovation systems for...It pays for the bureaucrats to innovate. The Presidential Hackathon is one. The innovative regulatory sandbox for another.
-
Basically, the idea is always the same. If you participate in, say, the Presidential Hackathon, your choice as the public servant is vast. You can innovate on any of those ideas. You don’t have to deliver it yourself. You can just partner with a civil society or a private sector friend who will then pitch for you. You can say, "Oh, I’m just their partner," or whatever.
-
If it fails, of course, you suffer no shame because you were just partnering. If it works then every year, we choose five teams. There’s no money in the prize. The money is the president’s promise that within the next budget year your idea will be part of public service. This is a hack to take the risk away but share the credit with the career public service. They love it.
-
The last Presidential Hackathon last year has more than 100 participants. All five out of five winning teams become every day operations in public service including using AI to detect water leakage, and using AI to prevent domestic abuse before they happen, and things like that.
-
These are the new story we want to tell is that we found ways to work around the silos in the ministries and cross the local and national regulations. Many such attempts are met with enthusiasm by younger career public servants. The cabinet office in Tokyo is also trying to do something like that lately.
-
I just visited Tokyo and met with people from the cabinet office. They say all the elderly states people are now giving more free reign to the young reformists in the career public service to try to come up with some kind of cross silo innovation reaching. I don’t know how well they’re going. At least there is public signals to that direction, which is new.
-
I saw that in theopen data index, Taiwan is on top.
-
Yes, but they stopped compiling that last year. We’ve been on top for two years while it’s there.
-
It´s on governments but it’s not comparable on enterprise side, right? Open data?
-
There’s the open data barometer, I think. No. It’s less easy to compare apple to apple when you have TSMC on one side and Foxconn on the other side. What does it even mean for them to be open data? There are data collaborative in Taiwan that stems from the private sector.
-
I don’t think it’s useful to have a global data index kind of thing for them. It may make sense to, for example, index them using the sustainable development goals to rate them on the impact of their programs on the society. Maybe we can do that but not on the raw number of data. I don’t think that works.
-
You want to share information. How far do private companies like the big ones, TSMC, Foxconn, cooperate with you, are willing to share their information?
-
That’s right but then there’s some call centers for them. For example, around cooling systems, around recirculation of water that are not their core business. They’re just they have to do as part of their business.
-
For these issues, they’re actually kind of willing to participate in the circular economy, panels and making sure that their technologies is also available for other people to use and so on. Of course, that’s not their core business. It’s the incidentals they have to produce. We see the same with the open source movement.
-
Still, most of AI cutting edge research is open source. That’s also because they hoard the data but not the algorithm. They’re perfectly happy to publish all the algorithm but not necessarily the data.
-
Every country has always the fear that too much information isiphoned off by somebody else.
-
That’s what I don’t understand. Unlike oil, when you copy data the original copy is still there. It’s additive. It’s intangible. You don’t take anything away by sharing data. You can’t siphon off data. The original copy is still there.
-
That’s because people connect AI directly to job loss instead of job reorganization or re architecting. That’s, again, why Taiwan’s missing based innovation model works. People can plainly see that you don’t lose job because of AI. You do have to re engineer your jobs to be co pilots with AI, essentially.
-
With less fear of job loss, there’s less cause to automation tax or universal basic income. That’s also a very popular response. The UBI movement in Taiwan is...There’s no pressing social tension for the UBI. UBI is like a litmus test. If someplace UBI has a huge advocacy, it means that there is some social tension around automation in particular sectors.
-
In Taiwan, because UBI never got very popular...I say this with full empathy. One of the main UBI proponent advocacy group is based right here. [laughs] That also means that we have almost no social tension around automation.
-
Taiwan’s geography makes it very clear that when we say something is a human right we deliver. Inequality can’t be felt in a bubbling way. If you go to any indigenous group or any rural places, you can plainly see that we’re pretty modern. Then the perception of AI taking away jobs is not easily sustained. People can clearly see it’s not the case.
-
You talk about all the social benefits.
-
Yes, like solving aging population issues and so on.
-
You need to have money to solve problems. If AI is too open, what’s the income base of the companies who pay their employees or the income base of the country to pay for all the social benefits?
-
It’s a risk and cause reducer. The question is, if electricity is freely available, how could company make money? Electricity enable a new kind of business. I think AI, it’s easier if you treat it less like a kind of new form of energy or a new form of current that’s run through all the different sectors.
-
As long as they speak data [laughs] they can be plugged into this new grid of automated or assistive intelligence. If you take this view, then it becomes obvious that you don’t focus on taxing the use of electricity. You can still do some of that, but you tax the surplus. You tax the outcome because of application of electricity.
-
It’s OK to be open about electricity standards. We center AI on the sockets, after all, [laughs] and it can be open around particular appliances, but not on the particular combination of the application, which is why people can very clearly see that when its core competence, then the data is siloed. If it’s not core competence, the data is shared.
-
The algorithm can always be shared. It doesn’t really matter. It’s a call center anyway.
-
What do you think about the idea that some kind of setup like a AI related technologies be taxedinstead of people? The more robots you have in a company, the more tax the company should pay.
-
It’s a interesting idea. It’s a interesting idea. As I said, it often reflects a social tension around job displacement by automation.
-
I think compared to UBI, of course, this is maybe more feasible...actually, on a similar scale. [laughs] In any case, yes, as I said, it’s a kind of regulatory response to social tension, because the representative really has to do something that convince their constituents.
-
In Taiwan, there’s as far as I can see no present pressure for the regulators and the policymakers to answer to the social tension, which is pretty low.
-
Again, can we talk about the question I started with? Where do you see Taiwan in 5 to 10 years as better off than right now, or is the competition getting too strong from other countries that Taiwan needs to fear...
-
Yeah, I think, first of all, Taiwan’s an island who erase five centimeters anyway, [laughs] whatever the humans do. [laughs] The biodiversity, the ecology would be there. But from a AI point of view, I think Taiwan is in a really good place when it comes to digital transformation.
-
AI Academy, people send their presentations. [laughs] Maybe I can just proxy in. I think they really have a good point here in saying that here, I think, of just...There’s really good examples around digital transformation.
-
Basically, what they do basically they’re a AI training camp. Their students are existing managers, like MBA as well as people part of their career is already around smart machinery or fabrics or whatever existing industries, and they are looking to transform their industries. That’s their constituents’, that’s their students’ space.
-
They basically work with lots of what we call smart machinery manufacturing or whatever companies and identify their pain points. Basically, there’s lots of university students, but they’re not readily helpful [laughs] when they want to digitally transform their industries.
-
What they are doing, basically, is focusing on the four particular use cases around flaw detection, predictive maintenance, automated flow control, and optimization around materials to basically prove that AI doesn’t need research. This is pure application.
-
On the pure application front, they don’t need to bring in outsiders. They can just digitally transform their existing workforce into people wielding AI, essentially, even if they are just a SME with 5 people or 10 people. That’s their main logic.
-
This is defect detection, and they made a point of not designing anything and sell as packaged solutions, but rather teach the industries how to co work with deep learning systems. These cases then turn themselves into advocates of this kind of co created optimizations and defect detection issues.
-
This one I think it’s called index. They needed to manually turn the parameters to get the quality consistent. After introducing a co pilot, they can turn the knobs for them. Then they can see that humans and AIs make different kinds of errors. That, basically, makes it clear that first, they go for their AI co piloting in their own industry.
-
Then they take these lessons away from their industry and become alums or teachers to the next batch who want to then spread this innovation to other SME industries and so on. It’s a rolling alum system. There’s huge amount of classes that’s going on. This is why I’m really optimistic in that you can see not only the code and the foundational infrastructure technologies, it’s democratized.
-
It’s the experience of integrating AI into miss mets that is being democratized. They can often take it to increase their job mobility for sure. It also make horizontal integration much easier than previously. They now all talk through the same language that is data.
-
I think in five years or so we’re going to see the miss mets still going strong. AI would have helped the horizontal integration that were previously impossible or very costly into general purpose partnerships not limited to within Taiwan but also internationally.
-
Awesome.
-
Right. Of course, you have to learn anyway. During learning, you have to make some essays and some testing anyway. Why don’t just make it as part of your digital transformation plan? That not only guarantee jobs but also guarantee a sense of money making.
-
Before AI it used to be difficult. In a highly automated pipeline, there is maybe only three neighboring people, a very large factory and in workplace isolation, loneliness, and so on. It’s a real problem. Now, all these parts could be automated. People can, again, go back to those central rooms of piloting remote construction workers [laughs] and so on.
-
They can be a social atmosphere and sharing. They’re essentially more strongly socializing functions. Now, they have to work with their community like researchers do instead of a purely automated kind of job.
-
It’s not only pushing people toward the creative or strategic dimension. It also pushing people toward a high socializing dimension as well. That’s a good vision to be guiding the development by.
-
Japan right now has a very new program for small children for learning AI related knowledge. Does Taiwan have similar program?
-
Yes. For using AI information literacy and so on, that is part of our curriculum starting from the seventh grade. The literacy like the ethics around data science and so on, it’s imbued in all the different classes as part of the information media literacy curriculum design criteria starting from the first grade, the primary school.
-
Then it’s part of every class. It’s not a class in itself. I think data science/programming/AI application it’s seventh grade.
-
On the 9th grade, there’s already lots of partnership classes with Google and Microsoft and also domestic AI training curriculum for people who already got a undergrad admission or people who are looking to work for a couple years before getting to undergrad to essentially equip themselves with AI over the summer or the second semester of the 10th grade. That is also very popular.
-
By the time they are in the undergrad, by next year, I think, half of Taiwanese undergrad student need to be capable of coding, regardless of their majors. Again, we fused coding into all existing majors rather than requiring everybody to learn "computer science."
-
I think that is also very, very integrated in the sense that if you’re a major in translation, of course you will learn about AI in the speech domain, in the text domain. If you’re a medical student, you learn about AI in the imaging domain, or expert analysis domain, and knowledge representation domain.
-
We’re not saying AI for AI’s sake, but enhancing the existing majors by incentivizing them to learn to code. That is pervasive in all our undergrads.
-
Graduate level, there’s four AI centers representing their region in Taiwan for AI based research. They coordinate their research programs that are more graduate study level. That’s the main shape of Taiwan’s AI integration in curriculum.
-
Mainly you talk about all the strong points of Taiwan. What would you say is the weak point of Taiwan on AI?
-
As I say, we’re not telling our story very well. It’s like the best kept secret. [laughs] It really is true. When I went abroad, many people, when they think about Taiwan, they still think the martial law days or the early days after the martial law, or the ’90s.
-
As we said, Taiwan wasn’t known for a flexible, innovative regulatory environment back then. Taiwan wasn’t really known actually for human rights back then. [laughs] Many people heard Taiwan is the first country for marriage equality to be realized constitutionally. They’re like, "What? Is this the Taiwan that I know about?"
-
That proves that we’ve been telling our story kind of terribly, actually, over the past 10 years while we’re being radically transformed into a by all means, pretty Switzerland like, regulatory system and human right system in Taiwan.
-
People’s catch, in their minds, are still kind of stagnant in the image maybe 20 years ago. I think that’s Taiwan’s main challenge, actually. We domestically also have generations that are used to dictatorship, or at least authoritarian thinking.
-
It’s a fault line. I’m the last generation to remember the martial law. Everybody younger takes freedom for granted, but everybody older is authoritarian. [laughs] I think this kind of culture clash, while not unique to Taiwan, really defines Taiwan politics, and defines many of the social controversies, protests, and so on, in the past few years.
-
We’ll eventually get over it, but we’re really facing a challenge in intergenerational solidarity. It’s really badly needed now. We’re not doing perfectly this.
-
Nobody’s perfect.
-
That’s right.
-
What are the best areas for German and Taiwan to cooperate in AI?
-
Lots of things. You can attend the presidential hackathon. We’re having a semi permanent relationship with New Zealand now, because they have a very similar program of three months of gov tech tech accelerator program, where the private sector, or social sector, can tell the government that, "This part need to be changed by AI, and I will show you how."
-
The government promises to implement the winning team’s idea into public service. In there, three months is right after our presidential hackathon. We’re just shuffling teams that won our hackathon to New Zealand.
-
It’s really a radical trust for the Wellington Water Company to be sharing all their SCADA data, like pressure measurement, or whatever, with the Taiwanese AI team. It flows pardon the pun both ways, because then we co create the solution around mitigating climate change, which is a new thing for everyone.
-
I think, around common topics, such as this, there’s less international competition because if you don’t mitigate climate change, nobody wins. For these kinds of issues, identify the Sustainable Development Goals, I think there’s natural synergy, natural partnership.
-
Really, if we don’t solve it, nobody’s a winner. On more purely economic issues, of course, the trade secrets and all those competitive issues you mentioned, will enter the picture.
-
My recommendation was just to identify the Sustainable Development Goals that you care the most. It could be around climate science. It could be around plastic waste in the ocean. It could be around upcycling existing agricultural materials.
-
It could be around anything, and then we apply AI to solve these common problems together.
-
Right. Thank you very much.
-
Thank you so much.
-
I hope it didn’t take too much of your time.
-
No, it’s just fine. I’ll send you the presentation from the AI Academy. I think it’s really high quality, worth a look. I’ll also send you the transcript for this conversation, and, of course, you will redact everything that you have said.
-
If there’s any snippets that you find that are worth quoting in its entirety, or you want to re write part of my sentence and so on, just let me know.