In 2014, our approval rate of the president at the time from the citizenry was just 9% and we specifically designed systems at the end of 2014 to start recovering some of that trust by asking people very simple questions: How do you feel about this controversial policy? Maybe it’s about Uber, so: How do you feel about people with no driver license picking up strangers they met on an app on their way to work and charging them for it? How do you feel? And of course, people feel very strongly. Some people…
…And I’m talking about the blockchain community, which is the zero or negative trust space. And they have now taken a norm that all speech that counts need to have like a zero-knowledge designated verifier proof and so on. And that changes the norm. So, everything else, including every tweet, is considered as a scam play. And only things that have the kind of designated proofs on chain and things like that are considered speech.
…each minister has trusted POs who bring their culture to a horizontal coalition, while each ministry leads its own open‑government plan. Recursively, agencies and departments have second‑level POs. Eventually, no agency fears engaging people directly on Join . That was the idea.
美國新任準總統川普任命「AI 及加密沙皇」（AI and Crypto Czar），近日在華府掀起熱烈討論。透過這項人事安排，可以看見美國 AI 戰略的重大轉變：不僅止於透過晶片出口管制等傳統手段防堵對手，而是更主動強化自身研發實力，加速在 AI 領域的領先進程。 12 月初，我出席了在華府舉行的「信賴科技峰會」（ Trusted Tech Summit），清楚感受到這股氛圍的轉變。與會專家普遍認為，美國若僅憑防堵策略，將難以長期維繫優勢。在此背景下，「進擊策略」呼之欲出：結合國際盟友，共同開發輔助研究的 AI 系統，真正達成「以 AI 研發 AI」。 這種「AI 科學家」的構想並非空談。OpenAI 近期公佈的 o3 模型正是重要指標：這款率先運用「審議式對齊」（deliberative alignment）訓練出的新世代 AI，不僅能擬定策略、找出正確思路後再輸出答案，更能根據政策與安全規範即時自我審視與調整，以達到在數學、物理、程式設計等高複雜度任務下，仍有強韌可信的表現。 在傳統的研發流程裡，打造一支能應對複雜挑戰的團隊，往往需要經年累月的訓練和經驗積累。如今，當 AI 進入研發流程，不但能在短時間內同時試行多條研究路線，也能迅速迭代實驗結果。如此一來，原本需要數年才能完成的艱鉅任務，或可在數月甚至數週內達成。這不僅縮短創新週期，更可帶動整條產業鏈同步升級，為尖端科技開啟前所未有的快速通道。 不過，無論是 AI 或加密技術，要在快速發展與安全穩定之間取得平衡都不容易，而這正是新任「沙皇」的關鍵任務。川普在選定 David Sacks…
…And so, I think a lot of people see that the blockchain space, because it is already maximally adversarial anyway, people invented like zero knowledge proofs and things like that, that protects the integrity of the information not due to any trust on any particular point, which may be DDoS or hijack or social engineering to oblivion, but rather into the resiliency of the protocol itself.
…The fabric of trust is very strong between urban and rural areas, across generations, genders, social classes, ethnic groups, and religions. And because we are not polarized along these lines, mere political polarization cannot turn us into a populist polity, which leaves very little room for populist leaders to gain influence.
…That exposed me to what we would now call second‑order cybernetics — trust , feedback, self‑organization. I later left middle school with my principal’s blessing after a science‑fair project using AI for philosophical inquiry, co-founded startups in Taiwan around mediatization and built intermediary algorithms to improve epistemic security and resilience . That entrepreneurial work eventually intersected with public service.
…In Taiwan, we say meeting face-to-face builds 30 percent of trust .
…So, it’s very important to me to understand when I’m collecting the database, to find with booking, with all the links that I got, if I can trust it, or there is a lot of problems.
…Also, around our own AI models, our science ministry is training the TAIDE, the trustworthy AI dialogue engine with the input, constitutional input of the people’s constitution of people’s online and offline participation. And the idea here is that people can participate in the AI model, people can simply specify what they expect of the AI model, and then through an assistive deliberation, we can merge all these different viewpoints into bridge-making narratives that can then be used to tune AI in a way that fits the societal norms. It centers on equitable…
…So, we do trust our citizens to also understand diplomatic and trade negotiation issues by being radically transparent of the functions and the factors that are involved in such conversations.
And so that helps to build trust . So I think so, two points here. One, compared to the previous paper based or other digital solution, it needs to be safer and also more convenient in order to get adoption. It’s not good enough to be safe but inconvenient, or convenient, but unsafe. It has to be better on both counts, first thing. And the second is that there’s always a need for people who feel like this could be better or this has done me wrong, to make concrete suggestions that can get…
…Okay? So the idea is not that we simply trust ChatGPT or Gemini or whatever to process all this data, but rather to make a smaller model that can run in the community infrastructure, and you can then verify that it has no connection at all to the outside world when it processes the data, but then it can still share the learnings through federation and so on to other plots, learning about things in the data without revealing the doxxing privacy information of the raw data. And so this kind of privacy technology is getting…
…However, there’s many, many other democratic countries going to be in the elections this year, and many of them are quite worried, frankly, speaking about the polarization that could result by this convincing synthetic content and so as part of the global Declaration, which chose the enhance the public trust , in fairness, enhancing transparency as the topics to consult with the people in Taiwan, about what exactly are the topics they care about? And are there any duties for the governments or for the large platforms are afforded from here AI labs that they would…
Yeah, I would recommend to trust the citizens more and to listen at scale. Those are my two main suggestions. For example, during COVID in Taiwan, in April 2020, there was a case where there’s a professional worker in a hostess bar, part of the nightlife district, that gets diagnosed. But she didn’t, initially, for the first day, disclose that she’s a professional working in the nightlife district because her code of ethics usually says that the anonymity of her patrons needs to be protected. While she worked with the medical office…
…And so we work on provenance technologies, works on pervasive FIDO rollout, and many other things, zero trust architecture, assuming breach, to make sure that the upcoming election in January stays safe from generative AI attacks. But as you know, it’s not just synthetic images and video and text, but also living off the land cybercrime that can synthesize zero day on the target’s computer. It also poses a potential biohazard. And five other things that I’m sure you’re very acutely aware of.
…And so by us becoming a competent authority, we can then shape it to say, you know, this is like a general pollution on the fabric of trust , and we can actually measure it on the effective polarization and so on. And so because of the negative externalities, we can either reassign that liability internally, like if somebody gets conned by a fake advertisement on Facebook, after Facebook refusing to take it down or to label it significantly, then that person may be conned for $1 million, and so we can reassign that and say, Facebook…
…That also means that she’s able to build trust with everyone in the room, which is one of the toughest tasks and one hell of a powerful skill for a leader.
…So unless it can be done through a trusted intermediary, which maintains contextual integrity and shares in a zero-knowledge way, nobody would want their open-source model and their fine-tunings to be measured in a way that could contribute to the risk sensing.
…But we think we need to look at the web3 world where there is like, you know, zero trust all around. And they have developed ways to cryptographically separate these two kinds of personal and non-personal information.
