 Welcome back to day three of the AI for Good Global Summit, and I'm joined now by Dario Tarraborelli, who's head of research at the Wikimedia Foundation. He gives us a new kind of a new angle on this, really. Obviously Wikipedia we all know about. So I guess it's really, you're here to talk about disseminating information. Correct, yes. So we're very excited to be here at this conference to talk about how AI and its implications for spreading information globally can be aligned with what we're doing at the Wikimedia Foundation. So Wikimedia Foundation is a non-profit behind Wikipedia. We operate some of the largest reference wards on the planet. And we're striving to build content that can be accessed by everyone on the planet, irrespective of their language, their demographic. And it is really critical for the success of our projects to have the best possible content accessible to everyone across the planet. So as a result of this, it is critical for us to engage with organizations that are trying to do the same and trying to provide learners and information seekers with the contents they need to become future citizens. And talking to all these experts on AI here, are you winning that battle about openness and sharing? I don't think we're winning that battle. I think it's going to be an interesting uphill battle. And the main reason is that what we see so far in the debate around AI is that we're still trying to figure out exactly how to deal with biases, with false positives, with a propagation of any errors and biases that AI basically amplifies at scale. And I think this is just the beginning of a conversation. And we do have some good lessons at Wikimedia that we learn by building AI ourselves in our systems. And these basically boil down to a few simple principles. Engaging with our communities from very early on, both at the design level, so what are the problems that we need to solve, how we design systems to solve these problems, but also in the creation of a label data and the evaluation of the model and the identification of false positives. Engaging these communities is really critical for building AI that is transparent and audible and inclusive. And a second angle that we focus on a lot is openness of the actual source code. So making it inspectable and audible so that we can learn from mistakes that we can make. Now I spoke to Professor at Berkeley yesterday. He said that robots will be able to read everything soon. Are the robots big fans of Wikimedia? Robots have been huge fans of Wikimedia. It turns out there's an aspect that is not very well known of Wikimedia. So we think of Wikimedia as an escalopedia for humans. Wikimedia is also the highest centrality node in the link data ecosystem. So basically provides a machine-readable data to pretty much every search-changing link data system that you can think of. And again, that makes it really important for us to think about what gets into the content of Wikipedia. Who writes Wikipedia? Because ultimately anything that gets into Wikipedia in terms of content gets propagated to the entire ecosystem of AI platforms and search platforms. And it's really important that we get the right content and recommend this contribute to it. Thanks very much. Absolutely. So that's Dario Taraberdi, head of research at the Wikimedia Foundation which I also have discovered has 300 employees focused on disseminating all that information now, every now and then. Thank you again. Thanks very much.