So, the previous stopgap measures against content/distribution doesn’t matter anymore, really, because we now have a very powerful rephrasing engine that can rephrase literally anything to speak to anyone with nuance.

Yeah, and I think you pointed out something very important: This is the actor/behavior level of things. This is not a content/distribution level of things, if we take the ABCD framework.

I imagine that quite a lot. It’s my job to do that.

I think a similar one was already also passed for non-consensual intimate images. And I think there’s another one about the digital double of election candidates that has also passed. So I think one of the main ideas is just to look at the immediate harms, but use the same ...

After we passed that law, Meta suddenly became much more responsible in Taiwan, so that anything that resembles a financial scam or things like that gets delisted, deplatformed or at least notified at the DNS RPZ in a very short time frame so that they have not been fined by ...

Yes. Here in Taiwan, if social media platform, or let’s just say Facebook, Facebook posts a deepfaked Audrey Tang recommending you to buy some stocks, not that it happened to me, but it did happen to other cabinet members. If Facebook does not notify the user that it is actually ...

Exactly.

Already in Taiwan, there are civic hackers using language models tuned by the civic hackers themselves to offer first kind of band-aid clarification to the spreading disinformation and so on, from the Cofacts project. So I think this empowerment of people closest to the pain plus some sort of liability ...

Once people become comfortable with finetuning and running their own models in service of them and their community, instead of those pretrained by large companies, they will be then able to develop that kind of antibodies, for example, clarifying the clogging and adding clarifying context.

So one thing we learned, of course, is that if we can assign, as early as possible, the liabilities of those centralized operators back to where it originates, right? Like the ongoing liability and accountability laws when it comes to social media harms, then this time we can just reassign ...

And they must not lag the top AI labs by a significant margin, because that was what happened a decade ago when social media deployed the narrow AI of persuasion technologies. The average person holding a phone wasn’t really aware that this is an addiction building machine. And they do ...

The main idea really is simple: This time we need to make sure that people closest to the pain, to the suffering, have the way to flag and seek redress and realign the AI models so that it empowers them, instead of concentrating power and taking power away from them.

Exactly. So, any solutions?

Yeah, it took literally centuries for us to teach corporations to care, right? ESG stuff, and even that is not really enforced.

It took, what, three centuries?

…precision cloggers.

As we know it, yes.

Isabel will be moderating.

It’s right before mine.

I think this is oolong tea.

Excellent.

So, like, averagely jet lagged?

So how’s the trip so far?

OK!

Ning Yeh is my vice minister in charge of legal affairs, for everything related to regulatory code.

Welcome, professor. We last met in 2004. Let me introduce Ning Yeh.

You too. Live long and prosper.

Looking forward to meet in even higher resolution.

I personally participate in many of the works of the CIP team, and I think when I say OpenAI and Anthropic seems to be broadly aligned, and that’s because we’re working to raise safety in those very short-term issues. So, I would encourage you to read a little bit on ...

I would like to add that our ministry is formally a partner of the Alignment Assemblies, which are convened by the Collective Intelligence Project, along with OpenAI and Anthropic, among others. I would encourage you to maybe read a little bit in the CIP blog.

Ours is January, so we will be the first pilot.

And if there are law amendments that need to be made, like in Taiwan, we already amended the law against generative AI in fraudulent election meddling, in revenge pornography, and also in scams, in financial scams. And these are the immediate harms that are raised and then immediately amended. So, ...

So, our idea, very simply said, is that no matter the imaginations are for a decade away, let us just quickly get the consensus around what the next 12 months we absolutely need to build a democratic assessment, assurance framework, threat indicators, and things like that. These are just common-sense ...

Instead of racing to optimize this or that feature, we need to race to safety, so that there is international coordination to tackle immediate harms to democratic processes, for example, to counter scams and convincing voice clones. These are whatever jurisdiction you ask, whether they’re in the progress camp or ...

Yeah, as you mentioned, the imaginations differ a little bit when you ask people what happens in 2030, for example, and that leads to different schools of thoughts in EU, UK, Japan and US and so on, for next year, which many jurisdictions will have elections, for the very short ...

Yes. So, this is what I call a level-two assistive intelligence. It’s not level-five autonomous in-silico conversation. We’re not there, and neither should we rush to get there. We should instead just take the best facilitators, the best citizen assemblies, run a really good face-to-face and online hybrid workshop, but ...

So, I think this is good for democracy, because it makes deliberative democracy more scalable, and also it makes AI development democratic, in that it’s not just a few engineers in the top labs deciding how it should behave, but rather the people themselves.

And now, in the face-to-face workshop, which lasts an entire day, we will make a long transcript of what people eventually agreed on as the norms, and this long context can then be used to train a constitutional AI model, so that this AI model behaves exactly the way the ...

Instead of a professional facilitator or moderator staying online all the time to answer people’s questions, this can easily scale up the deliberation process, making deliberative results accessible interactively to everyone who wants to participate in a conversation.

We can then use the detailed context as the agenda for face-to-face multi-stakeholder conversations, so that everybody can more holistically… it’s like a multi-dimensional view of an elephant, instead of each stakeholder feeling the elephant, different parts of the elephant in the room, we now have a multi-dimensional view that ...

So, this is what I mean by a democratic alignment. We’re just asking the citizens - how would you like generative AIs to behave? And then you can, through this, what we call a wiki survey, a survey with the survey questions written by the citizens, participants themselves, easily see ...

And then if you click into a cluster on the right, you will see the main arguments, which are also summarized by the reports on the top right. But if you click into a statement in the cluster, you should be able to have a conversation with that cluster. You ...