 I would like to first say that this will be a crowdsourced speech, meaning that whatever I'm saying is determined by your phone. If you can bring up your phone and scan the QR code, it's not a contact tracing QR code, it's a QR code for the online forum that can enable you to post any questions to me. And also like each other's questions, for I believe that the government and people in the healthcare community is now transitioning from working for the people to working with the people, and working with the people means that the people set the agenda of what you would like to hear me talk about. For the next 30 minutes or so, my speech will entirely be determined by the likes on this board, I see people are already giving likes. And the top questions, for example, currently, during the process you help government to fight the pandemic, what new experience does the task bring to you, and any new lessons learned you would like to share will be my first topic. And then the next most voted will be my second topic, third topic, answer one, two of the time runs out, is that okay with everyone? If you can't scan the QR code for some reason, you can also manually enter Slido.com and enter AskHIF, the capitalization doesn't matter, and then enter the same forum. Is that okay? Okay, awesome, so let's get started. In the past couple years, we've seen that Taiwan countered the pandemic with no lockdown and also countered the associated infodemic with no takedown. And these two is very important because that means that the people set the agenda, set the counter coronavirus measures, and the people also understood that public health science clarification is something that everyone is responsible of. For example, I mentioned the contact tracing QR code, the SMS, 192 to SMS QR code. That QR code is not my idea, it's not the idea of anyone in the government. Rather, it's from a community code of zero or G0V. This community started in 2012, systematically look at all the digital service that our government provides, which all ends in something that GOV, the government domain, and make better or at least more fun alternatives by changing an O to a zero. So if you don't like any particular digital service in Taiwan from the government, you do not have to go to the street for protesting or demonstration. Rather, the demonstration is in the form of a demo that is to say to simply build something that works better. It's the same for the mask rationing maps, for the self-service vaccine reservation system, also for the contact tracing system. So on this main, when a lot of people are worried that our previous paper-based contact tracing registration system in public venues are going to cause its own infection risk, because people share the same pen, or because people gather around in filling the forms of entering the venues and so on. So the leading architects of the private sector companies gather together in the GOV zero social sector online community to learn from the world what kind of contact tracing QR code system has been tried before, what are their drawbacks, and so on. And we finally settled on something that's a open standard across nations that's to say the QR code, and also an open standard across all telecommunication carriers that is the SMS system. Because if we invent new data collection endpoints that did not exist before the pandemic, people will naturally, reasonably, be wary of it, and also afraid of its privacy and cybersecurity implications. Which is why we did not invent any new data collection points. We simply reuse what's already there before the pandemic in a way that's very easy to explain. Moreover, people did not like the data being aggregated from the person or the place they trust, like their telecom company, to some third party that they do not trust or at least do not know as well as their telecom company. Which is why we said from the very beginning that of the quarter billion or so the SMS send since this May, all of it is stored in the telecom carrier itself and not shared even with the Ministry of Health and Welfare. It's only when a local outbreak happens, when the contact tracer needs this information, do they start to piece together the SMS from the five major telecoms to date? There's around 11 million out of the quarter billion or so that's been pulled by the local contact tracers in a multi-party way. And the second thing in addition to this crowd source innovation is the nature of the multi-party design. People understood that the 15 digit code is entirely random. That is to say the telecoms do not have a mean to know which venue the those 15 digit correspond to when you scan a contact tracing QR code. And just like piecing together a puzzle, that puzzle belongs to another QR code maker, actually there are three QR code makers, the trade van, CHT and the type of city government. They manufactured their own QR code. So individual privacy cannot be compromised if only one actor out of this multi-party like eight or nine party network suffers a security breach. None of them suffer security breach by the way. But where they suffer security breach, it will not compromise the privacy of the entire contact tracing system. So that's the second thing is a secure multi-party design. And the third thing is that it's entirely voluntary. So we're not forcing anyone to use the SMS, certainly not a QR code scanner. Just as I said that if you can't scan the QR code, there's a fallback, right? So people using feature phones, my grandma, for example, can manually enter the 15 digit QR code and SMS and text to already a trusted toll free number 1922 representing the Center for Disease Control, right? So that's a fallback. And for people who don't even have a phone with them, who don't have anyone accompany them, or they can still revert back to use the paper form. And well, they know that they'll not be crowded, they will not be shared by much because most people use the QR code. And then they also innovate it like using a stamp without the need for external ink with their name on it and so on. So you can just stamp your well into a venue which is arguably faster than a QR code set. Our innovations do not preclude future innovations. Anyone who want to make a new QR code scanner, and many people did, the Taiwan social distancing application is my preferred QR code scanner. But if you trust the trend micro more, you can use the trend micro QR code scanner. There's many QR code scanners by people because we based on open standard that's crossed and do not sacrifice any trust because we endorse a secure multiparty design. And I believe these principles, this is like a microcosm anecdote. These are the principles that will enable us to build more trust worthy data coalitions in the future based on the successful experience of the NIH Express app and the 1.2 to SMS contact tracing system. So the next question is in Mandarin. So should I speak Mandarin? I guess yes. The next question is, in your opinion, how should we prepare before the medical industry uses new technology? Instead, artificial intelligence can really fall into place and make medical health and our lives really improved. What I just said, the three principles are, the first one is the innovation of the open market. Anyone who uses it can, he thinks it's a bad part, he's innovation, not waiting for someone to help him innovation. This is the first one. The second one is, when this innovation really appears, we are not forced to put this data in what we don't believe in, but the data can always be put in what we believe in. That is, is this application layer to cooperate with the data flow flow, bring code to data, not the flow of the data to cooperate with the application layer, not bring data to code. This is the second thing I just said. This thing is actually a different power to artificial intelligence, to AI. For me, put this code to data, put the data in everyone's hands, or the place where you believe in, this is called assistive intelligence, which is the wisdom to assist. If you leave the data in one place, but you don't believe in it, or you don't know it, then I call it authoritarian intelligence, which is the wisdom to be authoritarian. What kind of value judgment does this have? It's just that the more people decide about what's going on, the more democratic it is. The less people decide on what's going on, the more authoritarian it is. This is not a very simple explanation of what value judgment is. When we were designing the AI system, of course, we could have chosen one of these two. How are we going to choose? This is actually very simple. I often say, assistive, but you might have done assistive technology. For example, I'm wearing this glasses, which is an assistive technology. Without it, I'm a bit differently able, right? But this glasses has two characteristics. The first one has value alignment, which means it's based on my best interest. I want to see it clearer. It helps me see it clearer. It doesn't replace me. It helps me. And I'm wearing it to see everyone. It won't suddenly appear in a 20-second ad, and I have to wait for it to turn off in 20 seconds. Because if it's like that, it's allowing it to somebody else, it's not allowing it to me. The second one is very important. It's accountability. When it's broken, it's easy to give an explanation. What does that mean? If I wear it today, and I find that it's loose, it's not stable, and so on, I go to YouTube to study, and I can fix it. Or I go to the blind spot I believe in, it doesn't have to pay a 200-million fee. It can easily fix it. It empowers end-users. For end-users, like what I just said, our contact tracing is based on what everyone can verify and what everyone understands. The QR code, the two open-source standards, the technology, and the eyes of the standard screw hole. Anyone can understand it, and it can fix it. So if you can do alignment and accountability, then you can prove to everyone that your artificial intelligence is an assistive, an assistive point of view. In that case, everyone can trust you and give you an explanation and be accountable and pass on your knowledge. So I think the so-called landing is what every user needs to know. Every user needs to know that I can understand this technology, that I can change it, that I can code to data. That means if I have a better place in this city, I can change it, and I can make it better. At this time, it's not your user, it's your partner. Or your collaborator. So the same goes for the more people turn into people with artificial intelligence, then at least in Taiwan, the more people can land, the more democratic artificial intelligence. I would like to answer this question. Eight people would like to know, where is Taiwan going in digital health in five years' time? What do you think are the biggest obstacles, and what do you think are some potential solutions? As I mentioned, we're now in a paradigm shift between the more Web 2.0 bringing data to code, to Web 3.0 bringing code to data. And it's not an easy sign to think in a way that is people-centric, mostly because it has been too easy for us to aggregate data in a single place. And it's rather difficult to share data in a secure, multi-party fashion until, well, last year or so, when there's many mathematical breakthroughs in secure, multi-party computation in homomorphic encryption, in which you can send the data in an encrypted form to somebody else to do the computation for you without revealing the raw data. But the computation after sending back, you can decrypt it and just review the computed results. So it doubles data controllership versus data processing. So that's another important breakthrough. Our National Center for High-Speed Computation just a couple of weeks ago run a conference with the support of some of you here to explore the cutting edge federated learning, secure, multi-party computation, homomorphic encryption, and things like that that will truly ensure this bring code to data approach. It's not just feasible in an academic sense, which has been proven as with some rigor in the past decade or so, but rather also viable in the economic sense in the sense that it's actually affordable and there's industrial standards and best practices that's already being formed around this new paradigm. So I think five years time, people will think it's normal to understand the algorithm, the code that affects their personal data sharing. And people will also understand that if they participate in the data coalition, just like participating in a local community building cooperative credit union and many other analog world data collisions, people will understand that they can continue to exercise control and also exercise voting rights and so on after they participated in the collaborative of the data. And we already see a lot of these prototypes nowadays starting to spring up thanks to the National Health Insurance Database and the prevalence of the NHI Express app, which I believe more than one quarter of the Taiwanese population are already using it actively. I think almost a third now, which is a kind of a critical mass for people to voluntarily contribute their data. We've seen, for example, last year when we announced that we're going to dedicate masks for international humanitarian aid. Well, more than 7 million masks are dedicated by the individuals using the NHI Express app, choosing instead of collecting their ration to mask quotas to dedicate it to international humanitarian community. So that's the Ministry of Foreign Affairs. It's not just acting for the people, but rather with the people, because we know in half of them, like the NFT choose to kind of embed their name in the open data on the list of wall of gratitude of people who dedicate their masks to international community, whereas half of them choose to remain anonymous. But one way or another, this is obviously already a working data coalition, although it's a smaller scale, but I believe we're now expanding it to many other areas in healthcare and also preventative medicine. The next question, what does it take for Taiwan to have open sourced health database for insight and for analysis on top of the NHI database, which is for academic usage mainly? From what I understand, if you bring code to data, then that is already the case. The director mentioned the chest X-rays, which is a good example. The chest X-ray algorithm, which is the result of international collaboration sourcing mainly from international chest ring, scanning data publicly available. It has trained a model that runs within the NIH database. That is to say, I believe in the Taipei and New Taipei municipalities, if you scan for chest rays and you upload to the NIH database as is the custom workflow, they will automatically be subject to this kind of rapid testing by chest X-ray AI model that runs within the NIH administration. And this pertains to what I just talked about, trusting the NIH, the virtual private network, that your data is the raw data, the scan. It's not going anywhere else. It's still within the NIH, which is tasked to protect and secure it. But the data is combined with code that is crowdsourced, that is contributed from the international health community and trust the NIH administration to run it within their data center. And of course, if multiple data centers in Taiwan or across the world choose to run this, then we can collaboratively train this AI model by sharing not the data, but rather the insight, right? The positives and negatives, the non-identifying information using differential privacy and many other mathematical constructs that will enable us to share what we have learned instead of the raw data on top of which, how we learned it because that would be sensitive private information. So I believe this sort of federated learning is already viable in Taiwan. It's actually part of our presidential hackathon in which every year our president choose five teams to win the trophy. And the trophy is the shape of Taiwan with a micro projector underneath. And if you turn on the projector, it projects the president giving you a trophy. So it's very meta, it describes itself. But this projection means that we're committed five teams every year to bring whatever they have innovated in the past three months or so to national level rollout. This is how we roll out the telemedicine as well as many other important regulatory changes with the help of Ministry of Health and Welfare and Ministry of Interior that enable the budgets, the personnel and the regulatory changes because the presidential hackathon process itself is a result of a slido-like what we call quadratic voting and participation from the social sector, private sector and the public sector to ensure its long-term viability. So yeah, feel free to continue to track the x-ray scan case with the NIH and I believe it shows the future. Without compromising anyone's privacy, everyone can nevertheless learn together. And I understand the next speakers in this forum will be talking about more details of this federated learning paradigm. Yo, Hu Weipo, when you say you think that the current technology and innovation can help Taiwan's hospitals, factories, and South China's cooperation to create a more simple exhibition and teaching training to help people with long-term medical care. That's a great question. Actually, what we just talked about, that bring code to data, we used to, the former president of the hackathon was the founder of my office in Shenzhen, in Shenzhen, Shenzhen. There are many new teams that are trying to do this. I remember there was a new team, a high-ranking new team, they tried to use AI to cut the waste box in Taiwan. And the example of this waste box is of course that the cost-effectiveness of the cost-effectiveness of Taiwan is not very high, but it has some pockets. So it's like what I just talked about, the example of the X-ray of the wall. The contribution of code and data to the whole world is made here in Taiwan. But in the end, we brought it to Ethiopia in the Social Enterprise World Forum to talk to a lot of friends in Africa about the WHO. In fact, one of our goals in 2030 is to eliminate waste boxes like waste boxes. But each scan has a lot of money. With the technology of Taiwan, it can be done in about a dollar. It can be done in about a dollar. This is very different to them. Although it's very different to us, because our health is very robust. It seems to be okay, but the situation in Africa is not like this. From their perspective, they are not very comfortable to say that all the raw data, all the source materials, all through the so-called cloud and all the information control are not like this in Taiwan. But because when we started training, it was a portable model. So they can use two stages of training there. After we train here, after we give it the model, they need to change the parameters of the model. Like we just talked about the way to learn in a similar way. So their data is different from Taiwan. It's kept in their hands. In fact, I think this is great. Because they don't need to absorb the biggest amount of training. They just need to use a smaller amount, a smaller amount of training. It can be done in a very important way for them. But of course, Africa is not like a country. I believe that you can also use any other jurisdiction in this country. Because they all have their own laws and so on. Bring code to data is not a problem. But if you bring data to code, it will be easy to touch the challenges we have across the country. I would like to answer this question. Five people would like to know is privacy still an issue of health data research in Taiwan? It's a motivational issue. It's not a problem to be solved. Rather it's a design constraint. It's privacy by default or privacy by design. As I mentioned during our Secure Multi-Party 192 to SMS-based contact tracing, the only reason why people would voluntarily use this new find-good system is because the individual pieces and components are well understood and people already understand that their telecom carriers have their telephone number anyway. So it's just like the SMS system. Now, a couple of weeks after rolling out this system and with more than 2 million individual venues voluntarily posting those QR codes there are some privacy issues and I'll be frank and honest about it. There was a judge I believe is his name that received a search warrant from a police investigation unit saying that they would like to access the database that turns those 15-digit random codes back to the venue because they are investigating an international fraud case or something like that and because through the regular wiretapping mechanism they receive a copy of the SMS of that criminal suspect and they couldn't make sense of it because it's randomized code so they would like the judge's help in issuing a search warrant. We've seen this actually happening in some countries that wrote out QR-based contact tracing system. In some countries the health office has actually asked police for data and in some jurisdictions the data has a law that says only for really serious crimes can the criminal investigation units share the data from the public health unit. But in Taiwan it's rather different because in each and every SMS that you send, if you read Mandarin, if you scan a 1922 QR code it'll attach to it that says this is for counter-epidemic use only. So this is basically the people's volition, the people's collective wish that this remains counter-pandemic only and which is why the CECC the Central Epidemic Command Center immediately started a consultation with leading legal scholars and also with the GovZero communities and very quickly we wrote out this public interpretation from the Ministry of Justice that says, well the normal communication is between two people and posting a posted note of 15 random digits in your telecom certainly does not communicate with anyone and this database is rotated every 28 days and that database for criminal investigation is rotated every 6 months so obviously this is not private communication and therefore if the telecom operators do not send a copy of the 1922 SMS to criminal investigation units it's entirely legal because it should not be wiretapped to begin with and so instead of like other jurisdictions where they try to kind of balance between sacrificing a little bit of liberty and privacy enabling a little bit of criminal investigation and so on we simply said, well this is for counter-epidemic use only and that is what won back people's trust on the system and then that if this SMS says for counter-epidemic use only it means exactly that and we make all the legal interpretations and so on public available so people understand we will not sacrifice privacy and imagine if we did not issue that interpretation imagine if people continue to worry about kind of sliding into compromise of privacy and so on then people will obviously stop scanning current code back to paper based stamping and so on and once that becomes the new norm well the entire QR code SMS counter tracing system fall apart as it did fall apart in many other jurisdictions so upholding privacy is the precondition of people participating voluntarily and trust is not just an abstract quantity but rather a continuing relationship that we have to continually earn trustworthiness and by trusting the people to contribute their data and by upholding their privacy to give no trust is to get no trust so to give trust to the citizens and their QR code scanning applications is to receive trustworthiness into the integrity of the privacy preserving nature of this when actually the SMS of this contact tracing design Yang Zhonglong would like to know if Taiwan does have the courage and opportunity to design a broad new national health insurance system say 3.0 how will it be in the future that's I think a question more suited for the two experts main table to answer because I'm the digital minister and digital is just an instrument a assistive technology to whatever public health goals and benefits that the national health insurance panel decided so for example I did help with the telemedicine system of QR code based telemedicine so that instead of the plastic IC car based NIH card anyone receiving telecare and telemedicine now in Taiwan can use their app the NHI express app and show a QR code to the camera and then to complete a transaction in the NHI database for diagnosis and so on and that of course necessitates a lot of change in the NHI express app but people overwhelmingly in our collaborative meeting and workshop said that it should work exactly the same as how we use a plastic based IC car they do not want for example mobile payments to be added to the NHI express app they do not want for example the public service usage that's not registered with the NHI express to start using the NHI express QR code as a proxy for electronic identification for example and so the people here in Taiwan have already a very strong know when it comes to the use of the NHI IC card that is to say the mutual accountability of the pharmacist or the clinician using their own card to register personal data transactions into interpersonal transactions that you can always account for any data mistyping or breaches and so on and so we need to carry that known forward if we are to still win people's trust in 192 to SMS case for example we built a website sms.192.gov.tw that anyone can enter their phone number and see in the past 28 days which municipalities or cities contact tracers with which number have accessed your SMS report and this is mutual accountability and this of course we learned from the NHI express app where from the from your private health bank you can see whichever information that your clinician or pharmacist told the NHI well you can also see it and also provide corrections if necessary so whichever 3.0 architecture that a new NHI adopts I believe around mutual accountability and citizen participation must still be the core of it another team called they are in the ambulance directly after the diagnosis through telemedicine next to the ambulance they can get ready immediately on the 2nd side can directly do some initial diagnosis and so on these are all very necessary different hospitals different departments use fire to confirm although they have different data source origin but when they switch they don't need to every point to write another point of exchange but we can all go to FHIR to do some kind of data hub to confirm that every person just need to write a fire adapter in a specific way a lot of details follow up on these teams next question is a new experience I have already shared it with English and the last one is how Taiwan should do to integrate and meaningful use of big healthcare data and I believe that data is never just personal it's always interpersonal and only through building data coalitions of strong mutual accountability and the freedom to remix at a very edge to empower people closest to the patient, close to the suffering can we truly build a data collaborative that is shared with all thank you very much