It’s not great for that. However, if you’re going to use it for scamming, for the con art, and so on, it’s perfect because it creates fake intimacy, fake trustworthiness , and so on. So, by the very nature of the newer generation of language models, it’s going to empower the people who are in it for criminal purposes more than the journalists who want to use it for fact checking, just because of the nature of the technology. So, it’s already asymmetrical from the get go.
Ambassador Audrey Tang is Taiwan’s Cyber Ambassador at large, and formerly its first digital minister. A self-described civic hacker Audrey helped transform Taiwan’s government into a global leader in digital democracy, using open data, participatory platforms and radical transparency to foster public trust . Now, as a fellow of the Accelerator Fellowship Programme here at Oxford, she is developing a project called ‘Plurality Advancing Ethical AI through Collective Intelligence’, a bold vision, and one that offers new ways of thinking about how societies can shape AI, not the other way.
First, but the public should be continuously invited to participate in a policymaking formulates mechanism and standards, which means that our aim is to at least run this kind of alignment assemblies twice a year, or maybe more as situations corporate large platforms should analyze and identify deceptive AI use that is difficult for humans to discern, which is almost all of them now and either choose just those are genuine, trustworthy and reliable to show the users or at least provide reliable verification and certification. For example, through digital signature mechanisms.
Underneath all of this is the AI evaluation system so that we can establish standards and application guidelines to ensure that the AI is trustworthy . Now, as you may also know, in addition to more than 300 startups with more than 10,000 employees, and that’s only counting the ones with good rank AI applications most like a downstream or was that already powered by them. These are built upon the foundations of the enterprise This is that produces the sauce tools, and of course also the AI chip and hardware.
Yeah, exactly. Let’s just get through the next crisis in the institutional phase that people have on democracy as an institution. But I do think if we keep coordinating globally instead of being, you know, the fabric of trust between countries being affected by deepfakes like the CFCs, Freons, and we heal the ozone, so to speak, through the international coordination, then that will lead the democracy to be much, much stronger because assistive language models running a decentralized way is actually very, very potent against autocracies. No autocracies like that.
And so that is the idea of interoperability. And we also need the idea of selective disclosure or meronymity, partial anonymity, because one part of sharing our lived experience is to identify the harm that is caused by those technologies. But if we have to sign each report with our real name, we're essentially doxing ourselves. And most people who are in a vulnerable situation would not want to contribute their lived experience to course correct those technologies simply because they would not trust , right, the people who process those data.
So, when it comes to implementing what we call digital public infrastructures or DPIs, we are more nimble. We can switch to passwordlessness quicker. We can implement zero trust quicker. We can do a comprehensive FidO rollout nationally. We can do a lot of AI evaluation and certification. And in fact, my ministry now also has deployed internally generative AI models for testing and just daily use, chat among the public service and so on. And so, because we move faster, we also discover more dangers or risk or issues and so on.
Yeah, so I’m mostly excited about the fact that it’s now possible for people of different cultures to build this kind of transcultural bridges by machine translation, translating not just across languages, but across, say, political ideologies, political divides, people of different backgrounds and so on. What used to take a lot of efforts in just building the basic rapport, right, basic trust between people of very different ideologies and lived experiences, suddenly can be entirely almost assistive intelligence automated by people investing enough in this kind of bridge making mechanisms.
Now, if we simply ask people to follow the rules, without publishing any information about the actual distribution of masks, that will be your way. That will be very top down. Instead, what we’re doing is that we trust these citizens with open data so that everyone can see every 30 seconds, which pharmacy’s near them, how much inventory they have in adults mask and children’s mask. The best thing is that this interface is not a government technology. It’s just civic technology created by people in Thailand city.
On Mobile Edge, the Microsoft mobile app, there is an integration. It’s not as user friendly as the browser extension that we built, but it does show that it can be done. We’re working with a new social media company based in the UK and a new search company based in Silicon Valley, run by a former top Google executive, Sridhar Ramaswamy. His search engine, called Neeeva, will integrate NewsGuard ratings and labels into search results so that people get instant guidance on the trustworthiness of sources in the search results.
Exactly, exactly. Yeah, so, for example, the zero- trust FidO implementation, of course, there are vendors in the market. But for Taiwan, our Ministry of Interior provides this TW FidO, the national implementation that includes digital signature and verification. So basically, it’s the same as what other countries use plastic eID cards do, in the form of an app, but this app is free of charge and it’s a public key infrastructure. Those reference implementations are public code, that’s to say, open source so people can modify it as they want.
And this year, although we saw comparable DDoS attacks and other assaults, it did not achieve the kind of results, as you said, last August anymore, because, as I mentioned, we switched to zero- trust defense systems, as well as asking people to help keeping us afloat, right, by making sure that we use the content distribution networks and also IPFS, which is a way for anyone, even people in regimes that are not democratic, if they want, they can run IPFS to help backing us up using their own hard disk and network.
…The ROOST is something that I am a trustee of. We believe that any moderation standards belong in the communities. It should be the families, the communities, that set the terms of service, not the big tech to set the terms of service. And we also believe that if any content moderation needs to be done, it's in a very specific scope, not by the big tech in a universal scope.
…We held alignment assemblies asking people how our AI affecting you, how do you feel about AI and with facilitated conversations, we actually tuned our sovereign model, the Taiwan Trustworthy AI Dialogue Engine (TAIDE) with the hopes and fears of people’s ideas in Taipei and Tainan. It turns out they have very different expectations about the AI’s role in the community, and we use so-called constitutional AI to tune the TAIDE to work people’s wishes.
…Only when people can verify that the summarization works in a way that people can audit repeatedly instead of sending all your data to mainframe and just trusting the mainframe operators vertically in super data centers doing the churning of the mathematics. Nowadays, people can download those models with the same downloadable data. As in Bowling Green, you can go to whatcouldbgbe.com and download all the data and then run exactly the same algorithm and see for yourself whether that summarization is biased or not.
And finally, we want to, as quickly as possible, switch all the places in our government and critical infrastructure that has national personal data to switch that entirely into zero trust architecture that makes sure that, even if the attacker takes over, my biometrics or my device, they cannot fully simulate my behavior pattern, because maybe I’ll be the target of the next spear-phishing, and my account may be compelled to do things that I wouldn’t usually do, right? And in those events, it will have to stop my account from doing…
…But I think the most important advertisement is really up to each ministry, because once one ministry like the Ministry of Health and Welfare which did really well with the first two petition cases, each petition case they handle well they create trust for those 5,000 people, and those 5,000 people, after they have another policy contribution, they’ll be like, "okay, they handled our petition very well, so now I’m motivated to propose even more interesting or even more radical proposals." I think word of mouth is really the first thing we…
…This not only hampers efforts to mitigate the impact of crises but also deepens societal fragmentation. Once social trust erodes, external forces find it easier to exploit vulnerabilities during major disasters, sowing further discord. How, then, can Taiwan proactively build social resilience? First, we must recognize that modern societies have shifted from vertical to horizontal trust . In the past, people readily accepted authoritative information from governments, academia, or select news outlets. Over the past decade, however, this dynamic has changed. Citizens are less inclined to trust top-down authority and…
…If we trust people, even 11-year-olds, with the ability to set the direction of their community online, they can actually mature up and grow up very quickly.
…Furthermore, the AI Evaluation Center also plans to assist the private sector in establishing AI testing laboratories, promoting domestic AI evaluation technology and promoting the development of Taiwan’s AI industry, thereby enhancing the trust AI environment in Taiwan.
