2020-11-26 Conversation with Microsoft OCTO

Greetings. How’s it going?

Awesome.

You’re really saying OCTO as “octo”? That’s great. [laughs] OK.

Awesome. I’ve read the briefing. Let’s go straight to the point. I think the briefing is great. In Taiwan, we don’t say “civil sector” that much. We say “social sector.” Otherwise, the terminology is the same.

I’m pretty sure that I know the work streams. Just for your situation awareness, the Microsoft Taiwan folks are also visiting – let me count – two and a half hours from now to talk about this particular meeting with the senior leadership. They will probably fill in on what their interpretation is for those work streams, but we did go through them briefly.

Thank you.

Yeah, sure. Yes, of course. Sure.

Yeah, a few things about the ordering of things. Usually, I will have the conversation around the COVID first, but I understand that your interest is mainly in the collective decision-making. Still, I think it will help if you reorder to get the infodemic one first, then the data coalitions, then collective decision-making, and then cross-sectoral partnerships.

The reason why is that the idea of infodemic is easy to make the connection to. If you only talk about misinformation, it seems like talking about spam, as in individual spam messages. If you talk about the infodemic, then it concentrates on the memetic nature of these things.

That’s an easier segue than starting just upfront with collective decision-making because collective decision-making means many different thing to many people. We want it to focus on the emerging issues if nobody has a clue or a holistic understanding without augmented collective intelligence, like the killer apps for augmented collective intelligence.

Otherwise, there are small-scale decisions that are perfectly fine using traditional methods. Just setting the scene, I think two, three, one, four, that’s the better sequence. Also, while we are on the word choices, you note that I use infodemic rather than misinformation. I will also note on the previous page. I think augmented collective intelligence or ACI is excellent.

I think the acronym if you put it somewhere, like on the second point on the last word, augmented intelligence (ACI), that’s great because first, I think it has a good parallel to the AGI stuff. Also, it makes us less likely to say instead, like you just did, collective augmented intelligence, which sounds a little bit more like Matrix or Soylent Green.

Augmented Collective Intelligence is just fine. That’s my only two observations.

If you would like specifics on the data coalition part, I have two more examples. These are, of course, Presidential Hackathon champions of this year. They are in the Presidential Hackathon page. I can get you the contact to the people running the show so you don’t have to learn from researchers or something.

They are all very eager because they just won the trophy to [laughs] realize their ideas. I will give the names in Mandarin, but there are also English subtitles to whatever pitch decks they did, I think. Here is the Mandarin thing. All of them are data coalitions. Just a second. Let’s see.

I have specific in mind a second link. There are five. They are all data coalitions. Maybe, I should describe very briefly on the second link what they do, and you can choose one or two that you would like me to highlight.

The first one uses augmented reality to simulate what will happen if we plant trees in all the unused public-owned spaces and the Internet of trees and forests and what the kind of environmental positive impact it could to get the data coalitions around gardeners, planters, and things like that. It’s called by Patch by Planting.

The second one is a Pokémon Go-like game to get people to go out and refill their water bottles instead of buying new plastic bottles. They can redeem their coins for local agricultural goods and so on. It also doubles as a push notification for heat damage potential so you can remind yourself to drink water.

The third one is using smart meters to train a deep learning network to learn about the specific behavior of home appliances so that you can save energy by replacing or augmenting those appliances without any intrusive in-house installations.

The fourth one is about getting workers and other frontline workers, especially agricultural workers and the families that care with them. Again, this is a heat damage thing, but also it uses the Health Bank or National Health Insurance Databank on a privacy-preserving computation to get the likelihood of people who suffered the most from the environmental hazards.

The push notifications they send is delivered to multiple apps including the tea-serving one, the bottle refilling one that I just mentioned to you.

The last one is about reporting illegal factories in agricultural lands. This builds upon the previous work of the WaterBox which is about water pollution monitoring in a participatory way, like AirBox but easier to model. This one in the supermarkets and convenience store, you can use their app to scan the barcode. For example, a box cookies.

Instead of displaying its prices, this displays its environmental impact, and maybe in the future, social impact so that people can use more informed choices as well as to push for the Ministry of Economic Affairs to publish the GPS coordinates of all the environmental fines and environment penalties so that they can feed that back into the data coalition model of the water pollution and things like that.

I guess all five are qualifying as augmented collective intelligence, and it’s up to you to do whatever. Their pitch films are there, and I think there are subtitles. I will just stop here.

Yeah.

Sure. There’s a single-payer health insurance system in Taiwan. It’s already recording all the insurance data in the Health Bank system, in the National Health Insurance Administration.

Anyone can use the app, which is the same app that you use to dedicate the masks to international humanitarian workers or preorders your masks to look at all the prescriptions and also X-ray and maybe CT scans also nowadays.

You can download it to your phone for analysis or use some SDK or things like that to authorize third-party applications to do analysis on those. It also provides privacy-preserving analysis possibilities for people who want to do exploratory computation.

Prof. Shih-Chun Candice Lung, a researcher from our national academy did this kind of longitudinal study, which correlates the factors of heat damage to not only temperature but also I guess the angle of the sun, wind, and things like that.

There’s a few factors in her research, and she correlated that with the Health Bank data without compromising privacy of the individual citizens.

Yeah, of course. You can download, of course, everything about yourself from the MyData portal, which is the blue button thing. You can also opt in to additional data collection and things like that simply using the National Health Insurance app, which I think has more than four million or maybe five million users nowadays, so a significant fraction of the population.

The NHI just released the de-identified data of everyone who has passed away because they don’t have – under the current legal framework – the right to privacy anymore. Anyway, these are raw data, de-identified and open for any research. If you want to do the research on live records, you can currently sign an NDA and do it old-fashioned enclave way.

We’re also looking into GAN-based synthetic data toolkits to get more data usability. If you want run it on the computational infrastructures that you do not directly control, we’re also looking into fully homomorphic encryption.

Yes.

Of course, I think bias is the symptom. The underlying cause, from our perspective, is the idea of equity in access to not only broadband but also media competence and digital competence.

If you detect bias, if everyone is equipped to detect bias or the bias on their friends and families’ prescriptions, and maybe it’s just an honest data collection error or whatever, and then they can just feed back to it.

Just like in the mask rationing map example, we did detect the systemic issues with the pharmacists handing out take-a-number systems and the idea of people using their ID card to collect on the spot, but we then just reconciled the two models to make sure that the pharmacists had finer control, not only on their opening schedule but also pressing a button to disappear from the map and things like that.

The point is that as long as the feedback cycle is open, as long as people have a way to just pick up the phone and call 1922 and get a more than 95 percent of not only getting their questions answered but also their demands met in timely fashion usually within 24 hours or at most a week, then bias is less of a problem because everybody is in this calculation together.

Yeah, my position statement on this, I just posted on the chatroom. I think the idea of cross-sectoral collaboration, which I would sometimes even put the social sector first, so it becomes the social sector first, then the business, and then the governance sector, means that the longer-lasting sectors have higher legitimacy.

The shorter-lived sectors – I don’t mean the government is neccessarily short-lived, but four-year terms and all that – should have less legitimacy by configuration, because otherwise, we risk these lower-on-legitimacy parties concentrating power to manufacture legitimacy out of thin air, kind of QE on legitimacy, citing for example the infodemic crisis.

As for your question, whether there’s challenges to this kind of work, I think it’s partly because we all – by we, I mean Taiwanese people above my age – remember martial law, and nobody want to go back to the censorship days or to the places where the media freedom and so on are challenged.

Basically, anything that starts with, “Maybe let’s just censor a little bit, takedown a little bit,” is a nonstarter because people will be like, “Oh, you prefer the White Terror?” That’s a quite unique innovation constraint, but that also makes it easier for the whole of society to be aligned to be basically not-PRC.

There’s really only one format question on the table. Because I always locally record everything, so I can just cut the pieces that I talked. There’s zero chance for people to gain knowledge into what your side of speech is, like zero-knowledge stuff.

Whether that is amenable to you or whether you would like the open discussion part to be both sides recorded and we weave them together, I’m fine either way.

That’s right. Yes.

Of course, because that’s entirely up to you. Radical transparency means at the root, meaning that it takes effort to take yourself out of the transcript. All you have to say is that you’re opting out.

Exactly, exactly. You will see only my side of the video and my side of the audio. This is better because in case of network connection issues and so on, you have the entire footage later on. You can – I don’t know – read my microexpressions using GPT-3 technology or something.

That point is that if you just say that upfront, then from that second onward, the transcript or the YouTube description will simply say, “This is local recording only because the SLT asked to opt out of their side of the recording,” and that’s it.

If you want to do it on the 1-minute mark, 16-minute mark, or 45-minutes mark, that’s your choice.

I am fine with just having 10 minutes because the four topics in the order you just listed, I only need 10 minutes for that.

Yeah, just share the deck to me. That’s the easiest. PowerPoint, built for sharing. [laughs]

That’s right.

I’m fine with that. The so-called internal discussion is simply just one-side-recorded discussions. I am fine. I will make sure that if you don’t want to be quoted, like if you don’t want me to reiterate your specific question because it’s a trade secret or something, I will make sure not to do so. You can see my recent Halifax Forum video to see that in practice.

Yeah, the Halifax Security Forum, not known for radical transparency, but I pasted it too. I’m pasting on the chat. Halifax is not known for radical transparency, but we managed to make that work, anyway.

Go ahead.

Yes, go ahead. Go ahead.

I wouldn’t say we work anywhere and everywhere, all the points in space time obviously. [laughs] The PRC, which stands for the People’s Republic of China regime, probably even if they are using Microsoft technology, it will take some time for this kind of augmented collective intelligence to work there.

So I wouldn’t say it like in a blanket statement. I will say two things, though. The first thing is that it’s been proven to work in Internet governance. The core norms of the Internet has not been broken. People have largely seen it in their interest, even the .cn and .ru folks, to still keep those ccTLDs online.

From the books coming out of the SLT, I understand that your faith is strong — but you needed proof. The idea here is simply to make the proof really solid. In the sense that it will be adapted to different forms everywhere else because everywhere you can find different tensions.

Our work is just to make those tensions, in a sense, grow upwing. In every place, there’s a left wing and right wing. If we can apply this X in RadicalxChange, the crossover ideas, that momentum always works, and that’s ideas worth spreading.

So for the idea that always spreads. I would, instead of saying digital democracy which is specific to democratic polities, I think augmented collective intelligence or ACI is a much more universal term for the kind of thing that we’re doing.

The “blah, blah, blah” part is right. We’re a very chaotic bunch, and we do get a lot of public opinions. In Taiwan’s case, I guess I will say two things. First is that there is a collective memory of martial law and a collective rejection of martial law. That is true.

I think we need to be honest and say that upfront, but I would also say that “rational self-interest for public good” is the main message that we’ve been spreading. For example, wear a mask to protect oneself from one’s own unwashed hands. Wear a mask because pink is the cool color, or some other rational self-interest messages. The same holds for countering the infodemic as well.

The second point is more to counter this “uniquely homogeneous” – I don’t know, Confucianism or whatever – part. Here I would highlight the fact that for the newer democracies, and especially successfully transitioned ones, people collectively don’t want to go back to non-democratic norms.

Thank you.

Thank you. I have a Cabinet meeting to attend. Sorry about that. [laughs]

Cheers. Live long and prosper. Bye.