So I’m really happy that TAIDE is a great example of how an AI developer, a group of AI developers use the large language models created by or offered by us for example, and to fine tune it with the data sets that are context specific and accurate. I think ...

I think I want to level set here the expectation, that the large foundational models are not necessarily serving as the end products and I continuously emphasized that the products developed, having a global audience as a potential user in the mind. I think I just want to make sure ...

I do want to ask. So go back to the topic to AIEC. I think the first question that asked in this meeting was about the AIEC evaluation metrics. And the only thing that differs AIEC from others is the traditional Chinese or Taiwanese Chinese accuracy.

As to Meta as an end user product provider. We are cautious in rolling out our products to the users. And I think that’s reflective of our approach is to make sure that they are created responsibility and with safety in mind. I think I’m happy to take the AIEC ...

I will be interested. I’m happy to see how accurate these are in terms of local language capabilities. But I would say that it’s foundational models so that is not meant to be served or to work as the end products for all. And I think one key takeaway from ...

so I think Meta has two aspects that would be relevant to this, one is a is Meta is a builder of certain large language models, and two Meta as a end-user product provider, the Meta AI that was mentioned earlier. So I would go first, the large language models. ...

I will share more information and I can share a link to the upcoming changes and system updates for you guys. But again, I think we are working with industry to improve our detection systems so that the vision of our platform is that when people are seeing content that ...

The system detection is continuing to evolve. And I would say let’s give them some time to build up a better detection system. The disclosure for the ads, it’s been there. But for the organic content, the disclosure products or tools or interfaces has not been created for now. That’s ...

So harmful contents will be removed based on our policies and the misinformation will be governed using our third party fact checking program. The third party fact checkers evaluate and to give rating and for repeated offenders that has continuously post misinformation that being ranked and rated by the fact ...

So again, as I mentioned earlier, that the contents that is created by generative AI are not necessarily deceptive and harmful. So deceptive, I think the definition of deception has to be pretty clear, that is done with bad intent on a malicious intent. And so if people are using ...

we are taking a very cautious approach. We do not believe this being rolled out to Taiwan anytime soon. There’ll be international expansion, but it’s going to be available in English language only. And yeah, so I think that people will still have access if they are traveling or they’re ...

So in terms of images, it’s definitely marked as made by AI. And that watermark, I think there are ongoing development to make sure that watermark is not just a visible watermark, but there are also like kind of DNA embedded in that. There’s technical aspect. I won’t go into ...

Yeah, so I had probably understood this question differently, but so let me just say first about my response to the questions that I perceived. So first about interaction. Meta does have AI tools for the end users. So in the US and in selected markets, if you use messenger ...

Alright, so I will share more information about the advertisements later because I know there are like public posts about that, and I can share that for the accuracy. But I just want to note here that we believe that transparency is important. So we built a system to detect. ...

So, at Meta as a social media company, we do think that we treat content with same standard, so AI generated content, labeling is one thing but even if it’s labeled as Gen AI, it doesn’t mean that they are harmful or so. Here’s a separate mechanism to evaluate or ...

And I think that requires a lot of investment as well. A good example was the industry standard about watermarking of the AI generated content, especially in video and images. The other thing is that because it’s open space, so that there are a lot of future models that can ...

So I think those are reflected in the conversation here as well. So, in our conversation, like people from from USA, from Brazil, from Germany, and from Spain also share the same concerns as Taiwanese do and some kinds of solutions. Such as labeling and citing the resources of information ...

So I think I just want to share this really briefly as a complement to the findings that we have here. Let me just pull out my note. The first finding from our survey from our community forum is that people are really genuinely excited about AI. And they are ...

We partner with Stanford University to run community forum on Metaverse governance. And that was a great experience. And so we also ran that in Taiwan. I think Taiwan set a clear example of how citizens can come together and discuss in a really informative way and make rational decisions, ...

Oh, thanks, Audrey. Well, thank you for having me to join this and I’m sorry I can’t be there in person and because I’m traveling and so but the technology enables us to join virtually. I think I want to congratulate these events and also the projects that Audrey and ...

Oh, actually, I have a question since the inception of the AIEC that we joined the opening last December I think. What I would believe that one of, as you just put an example there… It’s like focusing on the translation language accuracy. Are there going to be any standards ...