UA Day 2025 North America Regional Event: "Breaking Barriers: Universal Acceptance and Multilingualism as Gateways to Digital Access"

    1:02PM May 20, 2025

    Speakers:

    Keywords:

    Universal Acceptance

    Multilingualism

    Digital Access

    AI Translation

    Language Justice

    Internet Governance

    Domain Names

    Internationalized Domain Names

    Digital Divide

    Language Preservation

    Machine Translation

    Cultural Identity

    Internet Infrastructure

    Coalition on Digital Impact

    Language Diversity.

    Universal Acceptance

    Multilingualism

    Digital Access

    AI Systems

    Language Models

    Automatic Translation

    Multicultural Internet

    Language Preservation

    Education Technology

    Low Resource Languages

    Cultural Sensitivity

    Open Source

    Research Collaboration

    Stakeholder Engagement

    Future of AI.

    Question when it's time this is going to Be

    You? You

    Oh, good morning everyone, welcome to ICANN Washington DC Engagement Center.

    Recording in progress, okay,

    okay, I'll restart. Good morning. Everyone, welcome to ICANN, Washington, DC Engagement Center. I'm Matt Larson. I'm Vice President of Research in ICANN Office of the Chief Technology Officer, and I'm also managing director of this office. So we use the Engagement Center throughout the year to host events like this, to raise awareness around projects and topics that relate to ICANN mission to coordinate the internet system of unique identifiers. And today, we're happy to be hosting one of North America's universal acceptance day events. UA day events have been held throughout the year, since March and are ongoing. Let me go over some housekeeping stuff. Bathrooms are behind me. Go past the elevators and turn left, or you can go through that door. You're welcome to go outside on the terrace. It's a great day for it. The doors are locked, however, so please find an ICANN staff person with a fob, and they will let you out. Once you're out there, you will not be trapped. You can you can get back in without a key. You just can't get out. And then, in the very unlikely event of an emergency, ICANN staff will show you what to do. But the stair stairwells are also behind me past the elevators.

    So with that,

    my intro work is done, and I will turn it over to Christian Dawson. Christian is the executive director of the internet infrastructure coalition and co founder of the coalition on digital impact. So welcome Christian. Welcome again, and have a wonderful event.

    Thank you, Matt. I appreciate it.

    And yes, good morning everybody. To those of you who are joining us on this beautiful Washington DC day and those of you who are on live stream, thank you for being here with us today on behalf of it, coalition identity digital and the Internet Society and the newly forming Coalition on digital impact. It's my great pleasure to welcome you to this special celebration on universal acceptance day. I want to start by offering a heartfelt thanks to Matt and our friends and partners at ICANN, not only for helping us here today, but for their unwavering leadership over many, many years on the important subject of universal acceptance and a special thank you to the ICANN Board members who are joining us many over the live stream. We're glad to have you with us, and we are grateful for your continued commitment to this important work. Day is a celebration. It's also a call to action, because while the internet has made extraordinary strides in becoming more representative, still overwhelmingly Pharaoh favors, innate, oh my gosh, guys

    favors,

    I was a couple of minutes ago. I was running this over with my daughter, who's here, 15 year old daughter who's joining us here today. And I ran through this one sentence, and I said, I'm reading this thing, and I'm stumbling over words, and I stumbled over favors. I said pharaohs, and I just did it again. So now we've gotten that over with. We're good from here on Rowan, while the internet has made extraordinary strides in becoming more representative, it still overwhelmingly favors a narrow band of languages and scripts, and that's where universal acceptance comes in, ensuring that all valid domain names and email addresses work across all systems. Isn't just a technical effort, it's a human one. It's about building a digital world respects the linguistic and cultural identity of every person that logs on. It's about making sure that everyone everywhere can navigate the Internet in their own language. That's a principle that will guide much of what it is we're talking about here today, and the work that we are trying to do this principle, feels more urgent than ever right now, because, despite the fact that there are over 7000 languages in the world, over 5 billion people are online, only about 10 languages dominate most of the Internet content that's out there.

    Imbalance creates real

    inequity. Means most of the world's languages and the world views that they carry

    remain digitally invisible.

    That brings us to why now? Because just as we're making progress, we're also facing a turning point. It's cliche to say it at this point, but AI is changing everything, and it's changing it fast, without intentional action could make this imbalance worse, not better. Data that powers today's AI is overwhelmingly drawn from dominant data, rich languages, which means that AI shapes our digital experience, from search to translation to content generation based on those major languages. It's poised to reinforce the same exclusions we're trying to overcome. If we don't take intentional action, AI could accelerate the digital extinction of most of the world's languages with them, the culture cultures they carry if we don't take that definitive action, and that's why we're here today. I'm thrilled to announce the launch of a new coalition. We've been quietly building the coalition on digital impact Cody. Cody brings together a group of organizations dedicated to closing the language gap and addressing the deeper inequities it represents our founding members include identity, digital, Internet Society, public knowledge, American Library Association, the indigenous connectivity Institute, LGBT, tech dot Asia, your sky connections to coalition, Oda, HTTP, go, The Association of the internet industry, Black Knight solutions and CD and that's just to start together. We're committed to advancing multilingual and Universal Digital Access, from technical standards to public policy, from education to infrastructure and from content creation to cultural preservation. Cody's core mission is to bring awareness, advocacy and aligned action to this space. And we're not starting from scratch. We're building on the incredible work already done by ICANN UA steering group and leaders like Ron Mohan, my co founder of codi, who have long been the champions of digital access and digital inclusion. And yes, we'll be sharing most of a lot about what's coming next in a moment, including how you can get involved throughout the day, you're going to learn a lot more about codi websites now live at codi dot digital you'll find our founding vision a growing list of partners and opportunities to engage. Today's program will dig dig into some of the biggest questions we face. What does meaningful digital connection look like in a multilingual world? Will we ensure that AI becomes a tool for opportunity and not erasure? What will it take to build an internet where everyone, and not just English speakers and not just Latin script users, can fully participate? So I encourage you to lean in. This is a moment to listen, to share and connect.

    Let me close with this. The

    internet is one of the most powerful tools that's ever been built, unless it works for everyone in their own language, in their own voice. It isn't truly universal. We have the chance to change that with the right intention, the right partnerships and the right momentum, you can bring that vision to life for being here. Welcome to universal acceptance day.

    Welcome, my friend and colleague, co founder of codi, up here to take us through the first session. You're gonna be joined by Rebecca McKinnon, Rebecca, thank you so much for being here with us today. Would love I'm looking forward to the conversation you're gonna have. So much you

    the apps is working now, there we go. Good morning. Becca, it's so good to have you here, and I thought that perhaps we could start with you telling us a little bit of your story. Who are you and how did you get to Wikimedia.

    It's working now, well, to kind of backtrack or start from here and then go back and forward. I'm Rebecca McKinnon. I'm currently Vice President for Global advocacy at the Wikimedia Foundation. All of you certainly familiar with Wikipedia. You may have encountered the Wikimedia Foundation as well. We are the nonprofit organization that hosts Wikipedia and other related community run projects that are run and governed by volunteers in over 300 languages, which is, I think one of the reasons I'm here today. I've been with the Wikimedia Foundation for about three and a half years. Foundation itself is not a no, you know, I'm not an engineer or a product developer, but most of the 700 employees foundation or technical teams who are supporting volunteer contributors and being able to edit, contribute the knowledge to the projects and in as many languages as possible. Came to this job, very long journey that I'm not going to I really want to get to the meat of the conversation. But in 1004 I met Jimmy Wales when I was a fellow at the Berkman Center for Internet Society, now the Berkman Klein center, and we were so inspired by what he had done with Wikipedia, which was still in its early days, and its bottom up community governed nature that a colleague, Ethan Zuckerman and I co founded something called Global Voices, a citizen media organization that still exists today, and its most vibrant, the most vibrant Part of Global Voices at the moment is its language community and a program called Rising voices, run by a man from Bolivia named Eddie Avia, who were dedicated to supporting the revival of indigenous and threatened languages around the world, and so there's a really natural synergy. And I came originally from a journalism background. I spent a lot of time in China and got to know your work. Actually, I can, I guess it was 2009 to 2010 and ICANN was just putting into place the international main system, and how exciting that was. I was spending a lot of time with Chinese speaking communities, how exciting that was, and empowering was so many people. So I'm just really honored that coming full circle for today and be able to be part of what is really a movement in in promoting language diversity. That's That's

    great. So you and I were chatting, Rebecca, you were you were telling me about some projects that you've been involved in where Wikimedia is working on improving a content in underrepresented languages. Do you have any stories to share? Oh, lots of stories.

    So one, so, so, so, a few things. One, one story is just a project in Malaysia called wiki Katta, which was started by a bunch of university students wicking the Wikimedia movement has chapters all over the world. And there's, you know, a Malaysian user group and Indonesia chapter and so on, you know, all over. And so some of these students were contributing to Wikipedia in Malaysia and a grant to go out into the rainforest and help, you know, into the villages and launch a project to save this dying language called mandrig. And so they went around and and taught people in the village how to record pronunciations of basic vocabulary, and they would upload it and teaching people in the village how to do this. So it's not just people have to kind of parachute in, but teaching people in the village upload this to a sister project in Wikipedia called wikitionary, wikitonary, you know? And so that's kind of an online dictionary in all languages. So they would and basically upload enough lexicon for the language that begins to form the basis for machine translation and other things. And then they started teaching people in the village how to create stub Wikipedia articles about things in their community, in in their area, things that don't exist online about our people, their history, their culture, and in that way begin to create a foothold online, not only for their language itself, that can then feed into broader corpuses, but also just information about the fact that these people exist that really only is held by people who speak that language as their other tongue. And so that type of project is being duplicated all over the world,

    we're talking about revival, not just Preservation, not just preservation,

    but also revival, right? And and so helping to preserve dying languages also also ensuring that this language then becomes part of the corpus of structured data that can then eat into llms. And as you know, qpidia and its sister projects, it's pretty much you know, it's used by major source of training data for all llms Right now, and not just in English. And so the foundation is really prioritizing our language team has has developed a Translation Translation, a machine translation tool focuses on underserved languages and search support. So like as as a language begins to gain its foothold in the Wikipedia the media ecosystem might start with wikigenieri, or something called wiki source, where people can upload manuscripts, videos, Wikipedia, wiki Media Commons, where oral histories get get posted, and that is used as source material to build articles and and then eventually, an incubated language will become an efficient project over time, once it gets mature enough, as mature enough editing community, and then it really enters, kind of into the global Internet in a new way. And now one One thing to point out, too, we now have official Wikipedias or official hosted projects, whether it's Wikipedia or wikitery, Wikisource, 334 languages that's growing. 700 plus additional languages have test projects being given. So that's fantastic. It's growing. Now tell me this.

    I hear people telling me, hey, I have tools that just translate from one to another. Why do we actually need to worry about all these other underrepresented languages? All you need is tools that learn, translate, and the job is done. What are your views? Oh, there's a few problems with that,

    and and I can

    outside a particular example of that. So the there's a community, a user group in Kerala, in India, that focus focuses primarily on the language by help I'd always miss, right?

    It's a palindrome, exactly,

    and I always stumble on it. But so they make use of machine translation to translate, particularly stem information, you know, science and technology information, kind of globally consistent. You know, the periodic table of elements, whatever COVID 19 information into I am my all alone, but the information needs to be contextualized and localized. So there's one very prominent member of that community, Dr Nita Hussein, happens to be a medical doctor, and so she has been part of wicked project medical which is a independent nonprofit as well, making sure that as the COVID 19 and other scientific medical information is being translated into that language, that it's it's being done in a context that makes sense to The local communities and and that also will be information related to medical developments in the community that that are then written originally and and then and can be translated back. Um, also, there's a lot of information that that that community holds about its own people, its own culture, its own history, that just isn't in English, right? So you can't translate it. You have to translate it the other way. It has to go from my alum to into English. So it's a two way street. It can't just be lateral another, another example, just with that community, to dig deep into. One case study, is Pope Leo. So it happens that Kerala has a lot of Catholics, and so Pope Leo was great interest when that news broke, and the information about Pope Leo, apparently, kind of led drove more traffic to Wikipedia than ever before in history, but not just in English, in, you know, in all kinds of languages, including Malayalam. And see, I'm saying it enough. I'm finally getting there. So the local community used machine translation to translate the global information that was being contributed, not just in English, but in Italian and other languages. You know, Spanish, there was a lot of really relevant information that was coming from Spanish speaking communities and and then not only translating it into local language, but putting it in the context, in the order, like, what, what information is most important to that community? You know, he What connections might have been had to India and to Kerala in the past, etc, etc, and and so needs to be made relevant and put in a context that makes sense to people who speak that language. So, so it's it's just kind of translating from a generic English language thing, it's also coming from a context won't make sense to people in other contexts.

    Got it now, but what about say here in the US and North America where, you know, again, I hear voices that say English is the main language, and

    that's enough, yeah, um,

    the colleagues gave me some statistics that are really staggering. It's about 7000 languages around the world right now. Around 3000 could disappear in a generation. I someone from our community said that 20 languages disappear every week. I'm not sure kind of, you know, I can't cite the source for that, but, but these are the types of of, here's the get banding around and like, why should we care? It was at a Global Voices gathering where there were a lot of language activists who so contribute a great deal to Wikipedia, and a man, man named Jana Bata, who is involved with preserving this native language. Don't tell a which is a Nepalese language. He put it this way. He said, death of the language leads to the death of literature, culture and civilization, the death of heritage. Getting your mother language like drowning and disappearing. So just think about if you're enforcing language on rest of the world or drowning and disappearing, you are telling them they don't matter.

    Does that do?

    Oh, I think if we want everyone on the planet to thrive,

    to participate in

    their own economies, let alone the global economy, have opportunities to innovate, we if, if we believe that our young people in Nigeria deserve to be the next Mark Zuckerberg as much as anybody in California does,

    then this is important.

    If you tell most of the world they don't matter, their culture doesn't matter where they come from, doesn't matter they're setting ourselves up for ugliness. No, people don't respond well to that. I think a lot of the conflict we're seeing around the world leads to the fact that one people, one set of people, have told another set of people you are doesn't matter, right? And language speaks to the heart. That's, it's

    like you said it, it directly touches culture, heritage, art, all of that.

    What kind of

    if you had, you know, your the ability to just give a magic wand. What kind of a world would you? Would you have? Well,

    you know, I'm reminded of the ICANN tagline. I forget whether it's still in use, one world, one internet. I still pose an edit to that world, one internet, all the languages, right? Do? Love that we and in order that that is possible, and again, don't accomplish that top down. You accomplish that by bringing in the communities. And with Wikimedia, it's very important is that the tools we are developing the machine learning tools, the machine translation tools, the way in which we're working with AI, the way in which we interact with llms is we work for our communities. So it's very much CO developed in the communities, and that's very important here is we need to listen many of the young people. You know, I again at a lot of these gatherings that I've been attending, meet these Gen Z activists from Nigeria, Dr tochi precious might be online watching us right now from the Igbo community, or, you know, Ghana, Nepal, all across Latin America, etc, etc. Young people are just the amount of energy people have around bringing their communities online, in their languages, bringing their identities, who they are, their authentic selves, online, to be seen and heard and to have that opportunity. Just incredibly inspiring. You know, amidst all of craziness going on right now is one of the things that really energizes me to get up and open up that computer every day, is that these communities want to help with us, all of us, The IX we have, some of the founders of the internet system here want to work with us really achieve this, this vision, it's going to work for everybody. I got to do it now. I'm

    so inspired, and I hope all of you here and online are, I think, who are speaking about a movement, a movement rising, and it's so wonderful that wiki media is going to be part of it and help Cody get off The ground. So thank you so very much. Thank you. I

    now want to pivot to the next part. Thank you, Rebecca, to the next part of our conversation. I have Steve Crocker, who is an internet pioneer who helped shape early networking protocols, and he was a chair of ICANN when it first introduced the first internationalized domain names to the root of the internet. So I'd like to welcome Steve here. I'd also like to welcome Rachel Sterling. Rachel is the Chief Marketing Officer at identity digital. Rachel is focused on expanding impact and driving adoption of the company's 270 plus domain names portfolio. But Rachel also brings extensive senior leadership experience from Instagram, from Twitter and from Google. My mom

    is going to send you the $20 for saying nice mix about me. Thanks, Steve. It's so great that you're here. We're really excited that

    you're participating today. Well, thank you, and it's a pleasure to be here. Okay, great. Do you mind if I ask you a bunch of questions? I'm fully expecting. You're fully expecting. Okay, great.

    Let's hop into it. This is what we're referring to as lightning round or a lightning talk. Oh, thank you. But we're going to take as much time as we need, because what we're talking about is really important today. So obviously, Rob mentioned that you are a pioneer of the internet, the founder and greater ARPA ARPANET, CPIP protocols, and so when you were creating the internet. How did early Internet protocols contribute to global communication, and what limitations existed at that time in terms of linguistic inclusivity?

    Yeah. So this is where you get credit or blame. More of the latter. I think,

    I think we'll do a lot of credit. Yeah,

    so in some sense it's boring, and in some sense it's instructive. When the earliest part of the project was the ARPANET, which was all US based, almost, almost exclusively US based, and in a university and research environment predominantly, and the computer environment was that the computers of the day from different vendors were not, had not been designed and built to talk to each other. So you had, for example, the dominant computer manufacturer of the day was in a company that you hear very little about, called the IBM, and another one that was that I saw, I was joking a bit, another one that was quite important in the day, and has, in fact disappeared with Digital Equipment Corporation and but there were multiple other vendors, and the The internal representation of information was not identical across so just the way you would represent a letter, never mind a language or whatever, ABCD was coded differently inside these different computers ASCI emerged relatively soon after we started, but in the in the Beginning, we didn't even have a common way to represent letters and numbers, if you will. So

    all of these systems were communicating in isolation. There was no thought about how they would communicate with each other.

    That's that's the way they were they were designed. And so when we started to put the ARPANET together, which was to connect up these computers in different environments, we had to figure out what we could do that was common and what we couldn't do and leave a lot of room. And so the approach that we took was very, very ginger, in a sense, very, very delicate, in that we carved off thin layers of things that we could get to work together and then left for later development, and for multiple pathways to to multiple ways to to get it done, to what they would communicate at other levels. So we could move bits, you know, ones and zeros reliably from one place to another. And then when it came to moving letters and numbers, we had to an effect which, and one of the reasons why it's somewhat relevant to focus on this, it had a split effect. One effect was that, like computers would communicate very easily with each other, and we did not want to hobble them so that make them their their communication across the network less efficient just because they were in this common environment. And at the other on the other hand, when you wanted to communicate across these different kinds of systems, there had to be a way to translate from one to the other. And we actually wound up, if you look at for those who care about the details, if you look at, say, the File Transfer Protocol, which was one of the very earliest protocols, there was actually a switch in the protocol that said whether or not you were communicating in the native format, or whether or not you were communicating in a common translation format into ASCI and back out, for example. And that worked out pretty well, the sense that people got work done and they were able to develop applications, and it didn't put a strong focus on whether or not you had to be all one way or all the other way. That's kind of a, you know, the earliest signal that I that, I think one could pick out about the state of affairs that we're in now. So if you're communicating across multiple languages, and if you have English to English or French to French, then everything goes smoothly, and you don't want to impede that. And on the other hand, if you're trying to communicate English to French, or, you know, Chinese or something else, then you've got a more challenging situation, but but a necessary one you want to be able to deal with all that? So they so there's this sort of a picture from the earliest days that I think lays a repeated pattern for today. Long answer to your question, Oh, and

    if

    I may, I've made my entire career on the internet. Spent the last 18 years working in tech, and I am so overwhelmed being a chance to hear from you, because your innovation to create something that did not exist, what has powered so much commerce and has powered so much opportunity for everybody in this room, everybody beyond this room, and it's truly humbling to hear you speak about it. So thank you so much for that answer. Okay, let's get back to our question so I don't just geek out too much. What ways has modern technology succeeded or failed in bridging language gaps?

    Well, it's done a little bit, or maybe a lot even of both. First of all, the there's been extraordinary progress in understanding in computers engaging in

    human languages.

    So you have not

    only the written language, but you also have oral speaking. So we now have the ability to talk to a computer. This was science fiction back in the day. And if I might a little personal anecdote, I was privileged to work at DARPA in the early 70s after the ARPANET was started, but when it's still very early, days before the internet. And one of the projects which I inherited was the beginnings of a speech understanding project. So the technology that in those days, 1971 to pin down a date, was one word at a time. That was the available technology to talk to a computer. And the research project, which was super ambitious then and would be trivial today, was continuous speech, but in a very limited way, a limited vocabulary, a limited task, very clean environment, male speech, because it's easier than the higher pitched female voices. Whole series of parameters that made an approachable research project, and the goal was 1000 word vocabulary, which was huge, and I remember interacting directly with somebody on the staff in one of the congressional committees arguing about, how could we get people to speak fast enough to keep the computers busy? Because the computers were so expensive, and I'm serious, and I had to explain, you know, this was not real time, and we weren't trying to do it in real time. And he said, No, you don't understand. These things are really expensive. A crate computer is $500 an hour, and but this was indirect communication. I'm sitting at my desk in not too far away, in Roslyn, and he's somewhere in the bowels of Congress. And I thought, maybe I'll just go over and talk to him and and then I realized, did a little calculation of what his bandwidth was for paying attention to our research program compared to the budget that he had to deal with. And it came out to about 15 minutes a year. And I realized that in 15 minutes I could get in trouble, but I wouldn't have enough time to get out of trouble. And there. And then I realized there was a reason why a person in my position wasn't allowed to go over and talk to him directly. It was quite illuminating at the moment, the the the progress that's been made since then and now we, we talk to our computers and they talk back to us, and our our children approach these things as if, well, of course, they mean, it's a native language. Yeah, it's a native language. And they, if they go up to a to a device and it doesn't understand them. They don't understand what's wrong with the device. I mean, you know, just like they might try to scroll across a TV screen and say,

    I mean, when I when I ask my children if they want to watch a movie from before 2000 they ask if it's in black and

    white. And so, big generational divide, exactly, conceptual gap there. So the but the state of affairs now, very, very interesting. Well, I was ready, quite famous, well known professor at Carnegie Mellon University, and one of the pioneers in this area predicted and put money on the line as one of these traditional bets, that in 10 years, and this was a couple of years ago, 2021, I think by now, said, in 10 years, automatic translation between any pair of 100 languages in real time, so you'd be able to talk. And so, so that's pretty spectacular. When you sit back and you think about, well, to what extent does that mean if you're talking about technical things that are well that where all of the speakers are have the same background? Maybe that makes sense if you're talking about things that are more delicate, cultural or humanities oriented or poetry or so forth, the translation might be a much, much tougher and and then if you move up to things that are even more human about motivations and negotiation styles or interaction styles, you can Imagine trying to take somebody from a culture that is known to be in your face all the time. It can pick out one or two that you might think of versus others there where politeness and reticence are highly prized, and then trying to just merge these together in automatic translation. It might be a little more complicated.

    We were in Japan a few weeks ago, and Google Translate, provided that translation layer wasn't seamless, but we were getting there. But what I think is really inspiring about what you're talking about is the inclusion of culture as a cultural layer between high context versus low context cultures, and really being able to understand how you apply that layer to translation services, and so you can understand not just the words, but the meaning behind

    the words, right, right? And even meaning,

    in a literal sense, doesn't quite get at things like strategy or status and other elements, which are absolutely fundamental in human interaction. Instantaneously you know, you meet somebody in size, seconds less. And then

    it's crucial for us to understand how we incorporate that for all languages, not just data rich languages where business and commerce happen, but all languages on around the world.

    So so that gets at probably one of the crucial points for this meeting. What's the impact of the technology on languages? And I think things go in both directions. I think that dominant languages become more dominant because there's a natural what met cap calls a network effect, that everybody's doing it. And so it's a high benefit to being able to speak the dominant language. And then, on the other hand, you have the technology comes to the rescue of the less well populated languages, because it becomes pop possible to build a community, even if, even if it's a diaspora, and get all of the speakers and to get the corpus of that language and save languages from disappearing. That balance is and how it plays out, I don't actually have a strong prediction. It's easy, from a certain economic point of view, to say, oh, forget all these small languages. You know, one language will do it. Think was a lot of history that says that doesn't work, that a lot of resistance that. And on the other hand, you know how far you can go in preserving the smaller languages is going to depend a lot on the populations, on the on the human populations that are behind them, and who knows where that's going to

    go. Steve, thank you so much we are. We are probably going to get the AX right now, so even though we can talk about this for probably another hour, Christian is going to come on up and introduce our next round of

    speakers. Thank you very much. Real pleasure, real pleasure.

    Oh, I'm released. You're not released.

    Let's see i

    i see

    our speakers.

    So

    now we are going to do a panel on universal acceptance. Steve, thank you for continuing to join us and talk doing a little bit of a deep dive on universal acceptance and where we are with it. Today, I would like to add a couple of others for our panel. First, George Sadowsky,

    George, one

    of the many accomplishments that that George has in his history of working on the internet is to bring more than 50 developing countries online. Just incredible work in digital development, and so having his perspective on universal acceptance is going to be phenomenal. Also to add

    the panel I've got our only remote speaker. Think could be joining us.

    Magic.

    I see okay,

    in a moment, we'll be joined by Jeff, not Asia, and getting the perspective of of somebody working on this issue some aging communities is going to be also very vital for the conversation we want to have today. And by the way, while we're working on getting Jean online after this panel, we'll take a short break, so we're gonna get a little coffee so you have that to look forward to. A good moment. Thanks for being a big face on the screen behind me.

    Hi everyone. Sorry to be a gigantic, enormous head and happy to join remotely and currently on the other side of the world in Hong Kong.

    Everybody here? Jen, okay,

    okay, fantastic. So

    this is UA day, and one of the things we want to do in this panel is do a check in on universal acceptance, why we're still focused on it in 2025 what is left and.to be done. But I want to start those people that may not be familiar with the term, why it's important, by seeing if people can just give a little bit of a basic understanding of what we mean when we say universal acceptance. And I'm wondering if, again, you can start us out with that.

    Sure I'm gonna be the first one to talk about this, and hopefully you have heard this before. I'm really kind of cribbing from the material that I guess the universal acceptance steering group has. I hope you already know that as well. So universal acceptance really is the state or achieving. Universal acceptance is when we treat all valid domain names and email addresses, including the internationalized domain names, which we shorten as IDNs, and also long top level domains consistently. And that includes being able to accept them, being able to validate them, store them, process and display all of them as well. So that's in you know, in a nutshell, is what universal acceptance is. I can speak more about this, but I'm going to hand it right back to

    to you.

    So we're talking about a lot of technical work that happens at the DNS level to make sure that people can can utilize non Latin, not Latin scripts, scripts that are not just a traditional.com.org in the

    in internet usage,

    this stuff that at a technical level, correct me, if I'm wrong, has been mostly worked out. So why are we still talking about it in 2025 Oh George, let's, let's see if you turn that off.

    I think in part because it depends very much on incentives that that are very, very low or don't even exist in the part of people who run email systems and build web servers and the like. There isn't incentive if you're running an email system in the United States, there is not a lot of incentive to include or to change your systems to include a system which will handle 50 other languages, because most of your clientele is in English. So what we need to do is get systems that allow operators of these, of these input methods for for getting on the internet, the willingness to invest, in. It's not a major investment, but the willingness to take the trouble to go to universal acceptance and be recognized for it, we don't have it now.

    Incentives are off. Those

    incentives are, well, they're not economic because so they have to be something. They have to be either terms of prestige or contribute contribution to the common good. Okay,

    so let me,

    let me. Let me stick with you for a moment, George, because you work with a lot of developing countries, right? A lot of places that don't use prefer probably not to use the internet in English, right, right? Those also areas where there's potential

    markets.

    There may be no when I was working in those countries, there wasn't because the countries were barely scraping along on what they could import from the developed countries. But eventually what we should see, and I don't know if we've seen it or not, the pickup of UA tools in countries where the scripts are not Latin or Latin derivatives. I don't have an answer to that. Okay,

    George said it's an investment you have to convince people to make, but it's not a big investment. What are the big technical careers that people need to overcome in order to implement universal

    acceptance? Anybody

    not sure there are any that

    make the software just CP, my work, people

    aren't doing the work we have to do. We need to understand what

    the work is in

    just take a very, very narrow, specific example we've heard mentioned about internationalized domain names and the a lot of the software from the early days built in the wise assumption that top level domain names were Three letters, not four or five or six or two letters for a country code, and in many cases, that was hard coded in. And then when the use of longer names, this, even before getting to the internationalized domain names, came into vogue, the software just simply couldn't handle it. So there's been just an evolution to to remove some of those barriers that shouldn't have been there in the first place, but you can't control sort of the way software gets developed in many cases. So in a lot,

    in a lot of cases, what we're doing is we are, we are taking assumptions that were made a long time ago, that have been built upon and built upon and built upon, and trying to convince people to go back and redo their work on those early systems.

    Yes, and if the internet had been developed in Southeast Asia or India, or any place where there was a multiplicity of languages, this probably wouldn't have happened.

    So So Jen, you, you, you work with consumers in Asia, there is certainly probably a interest and demand in doing this work. What are the challenges that you face trying to convince people to do that work?

    Well, first of all, I think a lot of those users, especially in APAC, we're not native English speakers, so many of us may speak English as a second language, but given, I guess, from what George and also Steve mentioned about this history of how the Internet was developed, it really is means that the content, as well as the navigation of all of this is in English, mostly in English. I think I looked up the stats a little earlier, it's like over 50% of the content is in English, but free pack. That's not that's not the case. Actually, I think native and non English speakers only take up at about, I think, think it was 81% of the world population don't speak any English. So it's actually really hard to to actually fathom how, how different the internet could be as as I think that was a, you know, some decision. It was actually invented in Southeast Asia, just another parallel world. The crux of really is a lot of the users right now in APEC don't actually realize they can, don't realize that this technology actually exists because it is not being used in the platforms that they know and the apps that they actually use connect to the internet. So it's a little bit of a chicken and egg, but to actually have them understand this works, this also has to be taken up by, I guess, procurement processes, by the different ways that people actually connect to the internet, the different ways that people actually live their online lives. So it's, it's a little bit of a circular problem, but we have to start with somewhere. So talking to the APAC community, talking to the people who actually don't have English as their native language, they have to first understand that, oh, this actually exists, and hopefully in their jurisdiction, in their economies, that this will be taken up by different private sector as well as I think the biggest push there is to see that the different governments in impact. And there is actually really good example of Thailand as well, putting the procured process in there, in also the services they push out to their citizens. That is one way to get it started, so that that's really, you know, short of it, we really need to the semester across to the peoples that actually use internet, that they can use it in their different languages. But this also needs to be taken up, I guess, across a myriad of different actors, from government to private sector to people who actually change the codes, even for registries, registrars, we actually have roadmaps that we have to follow. You know, ICANN is a really good example to have a registrar, registry roadmap on how to get ourselves universal acceptance. Ready? But it takes, it takes a village, and it takes a global village to be able to make this happen.

    May it may take a coalition. Sorry, I had to George, you wanted something. Yeah.

    One can make the argument that the internet is really a follower in in using English as the as the major language the French have a term la Lang vehicular, the language that travels and and it's quite possible, in fact, in the business community and in the in the financial community, English is the language that travels. Fortunately, our brains can handle more than one language, and so I think we can handle both the English as as to speak to outside our group and like another language to speak inside the group. The internet may reverse that by virtue of universal acceptance, being more accepted and being being a reality, as along with the various machine translation tools that exist, we may be able to make and I think Rebecca commented on this, the ability to make small groups who speak a language viable, because you don't need economies of scale, and you can surmount this distance.

    I love that. One of the things that I took away from what you were talking about a moment ago, Jen was this idea. In a lot of cases, barrier to universal acceptance is sometimes just people's being able to imagine that it's possible to make that shift, and so building that awareness can help us avoid the outcome that Rebecca we were talking about before, where we have the possibility of of languages dying all the time and so much cultural erasure happening as a result.

    How? How can,

    how can we build this coalition in a way that gets all the right people focused on ensuring that they at least give let people know that these ideas are possible, so that they can apply pressure in in the right places.

    Really good point. And in fact, when, when Rebecca was talking about it earlier, I was, I was shocked and stunned at that. You know, the the stats that are coming up 20 languages signing every week, or is it every two weeks that's actually talking the first thing really is to know that, you know, we need that as a foundation of a multilingual internet, if we actually look at it in terms of UA by design, and that already shifts our thinking. And then the second thing is to step away. And of course, English is the language right now that travels, as George has said, it does travel, but stepping away from English first and moving towards more of a language justice movement, and this is the way that we get the shift in thinking both in terms of, I'm going to use that Asian example, how we, as a new technical operator, would look at it, and how others, especially government, there's the right pressure that you can apply to look at it in terms of UA first look at it in terms of, you're going to require this as A first step, we're going to ask in the procurement process, Hey, are you ready? And the next step would be, okay, we're going to look at those contractors that are UA ready, or have a UA readiness plan. And then finally, okay, we're only going to look at those that are actually, do, do have achieved universal acceptance. So it's a shift in thinking. And you know, you need people to start taking the first step, and the coalition is a really great first step here, again, couldn't help it. And I think this is really a great thing that we're doing right now.

    Appreciate it. Jen,

    Steve, we heard in your conversation with Rebecca a few moments ago that you were the chairman of ICANN, and ICANN made its commitments to focusing attention on universal acceptance. What is the proper role of internet governance? The I stars not just ICANN, but when it comes to

    universal acceptance,

    yeah, I'm glad you. I'm glad you made reference to I started, I get nervous when the reference made to ICANN and internet governance, as if those were the same thing. It's healthier all around, particularly healthy for ICANN not to be viewed as the governance mechanism for the entire internet in all aspects.

    Now the

    you know, I think there was common understanding as the internet grew that there had to be a embracing of the diversity of languages and cultures and a way to accommodate and not just tolerate, but positively and constructively accommodate them. So in the very specific and narrow issue of how to represent domain names, and not the content and websites, for example, but just the domain names. How to permit and support representations in scripts that were not the Latin script and not and not English based, there are combination of technical issues about, how do you represent those things? And there was some interesting, very, very interesting technical developments that made that efficient and possible for those of us who are into that kind of technical stuff. I found it was surprising some of the some of the work that was done, but the cultural and political elements of that were also very challenging and and quite interesting. For example, maybe a couple of examples, country codes. There was an original division made between top level domains representing countries, you know.de, for germany.ch, for Switzerland, and so forth, versus called generic top level domains, of which there were seven. And then it was expanded to sort of arbitrary number, and they were administered quite, quite separately. The country codes were left were administered by each country or territory, and the generic ones were administered under contract with ICANN of sort of kind of a split contractual and political environment. And then comes the introduction of additional scripts, and some confusion results because there are characters in different scripts that look the same across scripts that lead to some ambiguity, and that ambiguity can be exploited for negative purposes, for fooling people. And so you get an A character looks like an A, but is alpha in Greek or comparable in other scripts, and can fool people. So PayPal, for example, be a natural one, where it gets spelled wrong, and then people drawn into doing the wrong things. So comes a proposal from a country that says we would like not only our basic two letter code, but we want an additional one in a script which is natural for the language our our countrymen speak. And here's what we would like. And then when it comes time to look at sort of the holistic picture of it all, to say, Yeah, but if you choose that, it will look like something else.

    My favorite is,

    are you? Is for as the country code for Russia, you transliterate that into Cyrillic letters. Look like,

    pardon me, for, for the

    not exactly perfect. One

    could perceive this with the right eye, but it looks sort of like p y, not py, of course, it's, it's a Cyrillic for ru, but py is the country code for Paraguay. And so all of a sudden you get whether or not the IDN, the sort of first cut of the IDN for Russia, would would be confused with with Paraguay. Good. The Russians, to their credit, said, oh, about if we do Russian Federation equivalent, and so that becomes the equivalent of RF, and the the F becomes fine, like, like a fi in Greek. Thank you, thank you. And that is not confusable with any of the existing the Latin characters. So that's good news didn't hold across all of the others that put in proposals, and there were long delays while this got sorted out. Some of them got some of the interactions got to be quite heated and raised to political levels when you don't have a clean technical solution. And now we're in a very, very tiny limited I try to expand this to all of the other things that go into how do you have communication across these different things? Things get kind of complicated, so those are kind of things that get take up a lot of time. But once the course was set that we were in fact going to have support for other scripts so that we were going to support it, then one way or another, these things started to be sorted out. That's that's the path that was taken.

    It does sound tremendously complicated. We're very thankful for that work that you, that you've done to set the that has been done to to set us down this path that we then need to carry forward. We're carrying it forward into a generation with lots of new emerging technologies, some of which are going to make it even more complicated. What do you all think we talked a lot about what AI is going to do comes from universal acceptance. We had some hopefulness, but also some fears expressed before. You guys think

    they don't share your very optimistic view of AI, assuming, if it's very optimistic, I think it where it has paid off, just to the extent it is, resulted in massive abilities to translate one language to another. You know, many, many pairs, 1000s of pairs of languages. That's a major advance and and if we can be satisfied with translation that, I would say is good enough for most purposes, then I think we've got it today in major ways. So AI has delivered there. But, and it also depends what you mean by AI. Most people, I think, if you approached them on the street and said, What is AI, they would say, or they would they would think large language models, AI is a lot more than large language models.

    Let me offer

    related but slightly different point of view. I think we're in just even focusing on large language models. We're in a an early stage, not the actual beginning. That's maybe past us, but we're now have a widespread use of large language models. But nonetheless, there's a lot of evolution to go and we've been we are, have become very aware of the kind of errors it made from everything from making up specious references in in what are supposed to be a formal papers to just omissions and other kinds of errors. And I think the what's going to be interesting to watch is, how are these large language models and the use of them going to evolve as the errors are discovered and people either accommodate to them, or they apply pressure for higher standards as to what the accuracy is and whether or not you can actually use competing, large language models to challenge the correctness of what each of them are saying, for example, so you can have them, there may be some self Correction, and I think it's too early to know whether or not it's going to converge toward more accuracy and or whether it's just going to diverge, and we're going to have a mess on our hands. And I can't I wouldn't want to bet money either way, but I think that that's where the front is, if you will, in the emergence of all of this. To

    narrow my question on AI, before I throw to you, Jen, see if you have any thoughts on where we are with it there, and it comes to how people are interacting, we are generally talking about large language models when we're specifically in this case, there are these situations where is going to be easier for people to just go ahead and use one of the major languages, and so are we in a situation where some people could make the case at UA is they work is less relevant now, because it's easier for people who have no connection one of the one of the data rich languages, to go ahead and find a way that they can get by.

    Still muted. Thanks for narrowing the question. Christian, I appreciate that very much. In fact, I mean, you've, you've set me up really perfectly for the point I wanted to make. It's that without UA readiness, I guess the different languages, the multilingual nature of realities that we live in, we're gonna still remain second class citizens on the internet. If you have to do one more step back, actually do that translation, use that AI model to do that, that one extra step, it's still, still one step. So it is, again, as I've said, a shift in thinking. Of course, you know, I can't really, I don't have a crystal ball to see all the way down the line how this actually will look like, but knowing what we're trying to push for right now, I think it's still the correct way to say universal acceptance is very important. It is not not relevant, as long as we do have still the internet that we know, the domain name system that we know, the email systems that we know. Thanks Maria for also mentioning that it is still as relevant as it is right now and today.

    I think if you were advocating that because of AI and language translation, it's less important to have universal access, you're really sanctioning cultural literature.

    That's a good mic drop moment, right there. Okay, if we if we need a sound bite for today, that's a good one. So, okay, um, that being said, and I'm convinced that this work is important, that this work needs to continue to move forward. Right as we close out our panel today, I would love to hear from each of our panelists, what is one thing that the people who are motivated to take action and push forward work in universal acceptance to try to make sure move forward. What's one thing that they can do they leave here today to be an active part of this movement.

    Need to be able to we need to be able to make a difference, right? It's great to say, Oh yeah, I agree with this. But like, what can we do?

    I think we work on programs that that that encourage and, in fact, force, to some extent, acceptance of UA standards by people who are in a position to make them. I would love to see the consortium that's being established code into a country, and find some, some firm, some some service that has implemented the entire UA solidly and and publicize that give them an award and make available the fact that these are the people who can best represent all the language of the world in a way that is consistent and reliable and works. In other words, Praise, praise the people who go along, implicitly shaming the people who don't. That's great.

    And I think that what that does is it gets to what it is you were talking about before getting past this idea that Jen was talking about, where people can't even imagine

    not having

    to work in English, right? And if you go ahead and you publicize those examples, other places in which they have done, the work gets people past that imagination barrier. I love that, Steve.

    Let me take a combination of what both of you have said, and push it a step further.

    Picture emerges in my mind of a kind of a two layer structure in which every language group as a set of people who are advocates for inclusion, and all of the support necessary for that inclusion, that all of the tools are applicable and that the corpus of their language is made available and is actually used by training large language models and so forth and so for each of the many, many, many language groups, the emergence of those kind of support organizations, and they won't all happen uniformly at once. And then at a different level, there is, I think, a natural and very important role for the global organizations that are represented here, for ISOC, particularly for ICANN and your your organization, etc, to foster the that kind of formation within every language and for communication across them, so that the issues that are emerging get identified and raised to a level of visibility. So you say, Well, look, each of these groups is running into the following kind of problem. Is there a way to provide some sort of common support for that, or to research what the issues are and so forth in a more general way.

    Thank you. And Jennifer, what can people do? They want to

    active? Get motivated.

    I guess just what I said earlier, a shift in how you think about it, for tech companies, for big tech, for governments, shift towards the UA by design, for users to actually know that this is possible. And I guess Internet is a mirror of our society, and it should reflect the realities of this increasingly global, increasingly multilingual world. And you know, we really need to push this towards the language justice

    movement. And I will, I will end this out by going back to what it is you were talking about before around it takes a village where we're trying to build a little village, village over here with the coalition on digital impact. And we're we are eager for people to reach out, going to Cody that digital and finding out more about how it is your organization can get involved and be a part of that community. Was would be a great next step in enjoying the movement that we're talking about today. So I'm going to thank our panelists in just a moment before we head on to a quick coffee break. After the coffee break, we'll be back with our with our friend Vint Cerf here, who will join us to talk a little bit more about these important topics. But for now, I would like you to help me and thank my panelists on universal acceptance. You for those of you in the live stream, we'll be right Back.

    Recording stopped.

    Normally. Don't Think of thought

    This Is

    recording In Progress

    everyone. I'm Jon Perino, a Senior Policy and Advocacy expert at the Internet Society, and today I am pleased to be joined by a man who needs no introduction, but I'll give him one anyhow. And serve is known as one of the he's also known as one of the founders of the Internet Society, and our first president at the Internet Society. And you know, we're very glad to join, very excited to join this coalition, because the internet society's mission, our cause, is an internet for everyone, and to have that, we need a multilingual internet. So then I want to, I want to ask you about that. Back a bit thinking about ASCI, central internet. What were some of the challenges that you had to address and think about then, you know, ASCI is still around in some in some forms today, and challenges that have you seen since?

    Well, if you'll forgive me, I would like to spend a little bit of time on a kind of a preamble. It's not a peroration, although, as the Chief Internet Evangelist at Google, I've been known to sermonize from time to time. First of all, some of you might find it amusing to look at RFC number 20. I don't know whether you brought up Steve do. Okay, so RFC number 20 is written in October 16, 1969, October 16 is an important date for many of you who remember that ICANN, this Freedom Day occurs that point, many years later. But anyway, October 16, 1969 RFC 20 declared that ASCII was going to be the standard reference for expression in the ARPANET and subsequent internet was not envisioned at that time. So if you want to blame somebody for having chosen ASCII, it was me, or at least I documented it. That's the first point. The second point, though, is that when the world wide web came along in 1991

    Berners Lee

    simply chose to use Unicode as the way of representing characters in HTML. This could not have been more beneficial if you think about the world wide web as it is today and all the applications that run on it, because that allowed an enormous number of different scripts to be accommodated in the world wide web. So Tim needs to get credit for having recognized that ASCI was not the only thing that needed to be supported in his vision of the hypertext markup language. I want to also remind people who may be uncomfortable with the heavy element of Latin characters in the Internet environment that the Universal Postal Union, which precedes the ARPANET and the internet by decades, quite centuries, But certainly decades, chose to standardize on Latin characters for the names of countries, and this was to make sure that nothing else was understood on an address it was clear which country it was going to so there was precedence in the upu to Make use of ASCI, or effectively, Latin characters for this kind of standardization. I hope you don't mind. I was told that I was allowed to make and

    you are. You're a founder of the internet. You can do whatever you want today.

    Yeah, you know, when you get past 80, basically, you don't give a damn on anything. Do whatever you're going to do. So Steve brought up the ambiguity that was introduced with internationalized domain names. A specific example of the Russian Federation looking like Paraguay if you use the RU characters from Cyrillic. I wanted to also remind everyone that in other parts of the world where the language is expressed in other than ASCI, that there was actually a common decision to use ASCI for purposes of supporting the non ASCII language, and I'm thinking specifically opinion in China, which took some time. The inventor opinion had to sell this to Mao Zedong, and he had to try three times. For Mao Zedong was willing to accept the idea of entering characters in Latin script in order to refer to the phonetics of the character that you wanted in Chinese. So Pinyin was adopted as an easy way of introducing or getting to the Chinese script that was desired. Part of the reason for that is that the keyboards that were available were all in essentially Latin characters. Of course, there were variations. And as I understand it, there existed a Chinese typewriter which allowed you to enter each stroke. If you know anything about traditional Chinese, some of the characters have 50 strokes, and I can tell you, it just takes a hell of a long time to type that way. Of course, the Chinese users knew that too. So again, I want to try to illustrate that, even in our desire to be universally accessible, so to speak, that the ASCII characters or played a very important role

    and will continue to be

    so universal acceptance of all scripts and all representations looks attractive on the surface and it sounds egalitarian. It also, however, introduces potential ambiguity. Steve's example was one kind of an ambiguity, but there are others. Let's suppose that there are multiple ways of representing the same country code. Not everyone have facility with all representations. There are some people who will not be able to read the non Latin characters, and so when they are looking at a domain name or referring to an email, they may not know for sure exactly where the hell is going, because they don't know how to read that script if you have more than one representation, and it's not always clear which one is is correct, so you could conceivably mistakenly go where you didn't intend to go. So that's a potential hazard. One possibility, of course, would be to map all of the reference patients into one standard representation. And that's sometimes a way of getting rid of ambiguity. If everything goes to a canonical form. I spent two years of my life working on the ID and a 2008 which didn't emerge until 2010 and it was carefully designed so that the puny code representation, the ASCII representation of the non Latin domain names, would map uniformly and symmetrically Back and forth. So there was a guaranteed mapping from the one puny code form, which captured the non Latin character set to the to the

    Unicode form, Unicode being ASCI only and Punic, and the Unicode being in the very broad character set. Unfortunately, the parties who had already adopted an IDNA in 2003

    resisted of the restrictions of 2008 that enforce this, this symmetric symmetry, which we thought was important, because then you could assure that you were getting the domain name that you expected. So unfortunately, there is still ambiguity as a consequence of ID and a 2000 threes continued use. I want to take a moment, if you don't mind, to respond to the artificial intelligence discussions that we heard earlier with Steve and others. I am actually more positive now about large language models and their functionality than I was even a month ago. I'd had plenty of opportunity to try these things out. Google, as you know, has invested heavily in this space, our Gemini sequence being the primary example, although we have many other language models, some large and some small. What I have found extremely, just striking, is that we've used these technologies recognize languages and to automatically transcribe them and gotten to translation yet but transform specifically stunned recently, I was on YouTube watching a video of an Italian news press conference, which strikingly and oddly was in Italian. And as I was watching it, thinking, you know, yeah, no, speak of the language.

    A little

    window popped up with the

    English language translation of the Italian. I didn't ask for this. Had set my settings in Chrome say my preferred language was English, but I didn't ask for it to do translation. It did it automatically. It was not bad to the extent that I could recognize some of the Italian words. I sent a note to the president of YouTube, Bill Mohammed, and I said, I didn't know we could do that. And he said, Well, I don't know where you've been, but we've been doing it for a year. And I just did I hadn't watched any videos that were in other than English up until that point. So I have been quite stunned at the ability to do transcription and automatic language recognition and also translation. Moreover, it's reached the point now where we can do this on mobiles, entirely on the mobile itself without actually going into the internet. And so there's for trans transcription. In particular, there is an application called Live transcribe. It was designed and built by a Russian engineer, Dmitri kansky, who's deaf, and he needed this in order to work with his colleagues at Google, and so he developed this. It originally ran by going out to the internet and coming back again. Now it runs entirely on the mobiles, at least if you're running Android. And my wife uses it all the time. She has two cochlear implants. She finds it really helpful to see the text transcriptions. So I am anticipating today, which will not be long from now, when you'll be able to set a language on your mobile, set it down on the table and have a conversation with someone who is speaking other than your preferred language, and have it automatically translated. And of course, if you both have these things, you can both be translating at the same time. Here's another example of utility. Here imagine that you have a one of those foldable mobile so that you have two displays. Could be showing its output to two different people at the same time. Imagining you're a doctor talking to a patient who doesn't speak your language, they're seeing the translation. Both of you are seeing the translation. You see what you said in English, but you also see what the other person is seeing. If it's mistranslating, you can detect that, and you can make corrections on the fly to make sure you're communicating something. So that's my preamble to excitement about the introduction of these non Latin character representations, both within the World Wide Web and in the domain name system. I will say, though that the uptake of internationalized domain names has been less vigorous than many of us expected. I think some of the reasons for that are as I tried to describe earlier, like the opinion preference. Oh no, I leave it to you to guide me through the rest of my talk.

    Thank you. Thank you. Vint. You know, as a history buff. And when we think about ASCII, I mean, we're looking back at the graph right in some way. So I think that was actually a pretty short introduction of where we are from then to today. But something that still exists today is the digital divide. And when we think about the digital divide, oftentimes we think about the infrastructure. Think about the affordability, the reliability of an internet connection. There's this other part of this right, which is language. So then, what is the role and importance of language in closing this digital divide?

    I think we touch of it. I'm language disabled. I speak English. I have some German little phrases in Russian specific and just enough French to order the wine and find the men's room. So so I am language disabled, and I think that having technologies that can help us overcome that barrier, that efficiency is very important, and I think we are much farther along now than we ever were before. In closing that particular digital divide, one that I worry more about, to be quite honest with you, is kind of understanding of how you use these technologies, the kinds of hazards that we face when we get online are significant and growing, whether that's hacking or phishing or misinformation and disinformation, there's malware, there's a long list of potential hazards online, and some of those hazards can be exacerbated by not being able to validate what we're seeing, particularly if you don't speak the language of the message coming in, might ignore it, and it might be a good thing, because it was a phishing message and you didn't understand it, so you threw it away. On the other hand, it might have been a very important announcement that you should pay attention to. So our ability to translate correctly is important. So I see a digital line in many dimensions. They are economic. They are, I would say, knowledge of how to operate in an online environment. Some of you may have heard me argue for an internet driver's license. And you know, think I don't really think you have to be issued a license to use the internet, but think about what we insist on teenagers learning before they get a license to drive a car. Tell them you should make the assumption that every other driver is crazy except you, and you should learn how to drive defensively. You need to demonstrate passing a test that you know what the rules of the road are. Need to demonstrate that you know how to manage the car, park it and parallel and do all these other things. Um, and so, in a way, I think we should be teaching people similar kinds of things about the internet and its hazards and its intended use, people should be able to demonstrate their understanding of not arguing that we have to get a license to use the internet, but we should do the same kind of training, and that, I think, would help close some of the some aspects of the digital divide. Still have a lot of work to do on the economic side, making things affordable. And in an earlier discussion this morning with the Marconi society people, we were looking at internet resilience all of its dimensions, and keeping this thing reliably running is another important divide because it doesn't run reliably everywhere where it should. It's not affordable everywhere where it should be. Many, many other dimensions of that divide that need to be worked on. So the job is still open for us to close kinds of divides show up in our Internet environment that was such a long answer to a simple question.

    Thank you. Now, in the year 2025, I do have to ask you about artificial intelligence, or I believe I'd be booted out of an event. So tell me a little about AI machine learning. Obviously, there are opportunities. Already talked about what YouTube does in translation. It's not necessarily new. Also, some limits to that. Where does that fit into this conversation? There's some, some, some limits to

    power as well. I mean, it doesn't the artificial intelligence applications of today don't always work perfectly. Almost nothing ever works perfectly. It's getting better. I mean, I often tell an anecdote about a weakness called hallucination. Some of you heard this story. I decided to test one of the chat bots by asking to write an obituary for me. Some people think that's kind of sad, but you know, I figured that the system would have seen the format of an obituary, because people die every day and their obituaries appear on the internet. And if you're trained on the content of the World Wide Web, you see an obituary, and you know what the format is. And I figured there's at least some stuff on Wikipedia about me, so asking it to write an obituary seemed like a reasonable thing, and indeed, it started out the way you'd expect. It says, We're sorry to report Dr sir passed away, and it gave a date, which was much too soon. I didn't like that at all. And then it got to my career, and it conflated things I did things other people did. I got credit for stuff I didn't do. Other people got credit for stuff I did, and then they got to the family members part, and it made up family members that as far as I know I don't have. So I remember scratching my head thinking, How the hell did that happen? I mean, if this thing had been trained on factual information, how did it produce counter factual output? I'm not an expert in the implementation of AI, but I imagine in my little cartoon way, one reason for this kind of failure is that when the system is ingesting content, may not have enough context to understand that this paragraph, for example, Steve's bio and mind could easily show up on the same page. We've both been previous chairman said ICANN, we both been involved in ISOC we've worked on the internet. Steve's bio and my bio could be adjacent to each other on a web page. And while this thing is ingesting worldwide web content, it can be words from his bio and words from my bio and then generating what's the probability that this word would follow that word within a certain number of other words in a cartoony way, how some of the AI systems generate output. Here's a string. What's the next like, most likely word we put out in that string? I could see how the lack of context of these two paragraphs could cause it to conflate our BIOS and produce hallucinated output. I will say, though, that as I continue doing these tests, the systems are increasingly able to avoid those problems. Some of it comes from pre training. Some of it comes from having a much larger context window in which to generate the output. So in my experiments with the most recent of Google's language models, I haven't been able to induce the same kind of hallucination that I was able to do a year ago. So I'm becoming increasingly persuaded that these are going to be powerful, enabling people to use online resources. Our CEO Sundar estimated that about 30% of all the code that's being generated by Google is being generated by AI media. My first reaction was, I sure as hell hope there was a human being that looked at that software before it was introduced into operation. And indeed, there is a whole process for doing that. But nonetheless, I think we are just beginning to see an era where AI tools of various kinds with various specialties will be enabled. They will make people more productive. To give you one example that I read about recently, there was an you know, that we have this Google Deep Mind team. They were the ones that invented AlphaGo, that played the go game some years ago and won four games out of five against an international Grand Master. AlphaGo was succeeded by alpha fold. Alpha fold figured out how to shape, how to how to predict the shape of 200 million proteins that could be generated by interpreting the human DNA. So we have a big catalog now of what the proteins look like. You could use the catalog to figure out maybe not interfere with the disease process of COVID being a good example, a small molecule that attaches to the prong of the COVID virus interferes with its ability to dock with the ACE two receptor, and there inhibits its propagation. So we now have things, most recently, literally, on the 14th of May, just a few days ago, another report was released. It's in the artificial intelligence is called Alpha evolve. This is a really interesting piece of software, because what it does is generate programs to solve problems. It's not solving the problem, it's writing programs that can solve the problem. So it's a step away. It's like the difference between functions and mathematics and functionals. So they got this thing, they posed a problem, which was to make more efficient matrix multiplies, because that's how these large language models work. They do matrix multiplication. 56 years ago, somebody figured out that the most efficient way to do a two by two matrix multiply involved 49 multiplications, and that was it. For 56 years, nobody could find anything better. Alpha Evolve is set go and find a program that will help discover whether there is a smaller number of multiplies to do the same thing, it comes up with an answer. It gets 48 multiplies out of 49 now I'm sure you're not levitating out of your seat because we got one less multiply, but you have to remember that when we're running these models, we're doing billions and billions and billions of those multiplication is to exercise the model, either to train it or to do inference unless multiply significant in the context of scale, and this program figured it out. So I'm, you know, now, super excited about the capabilities of some of these systems at scale. No human being would ever be able to do. I'm much more enthusiastic about this stuff than I ever was before there.

    Thank you, Vin. I have one last question for you, and as you just said, there's incredible opportunity right now. I think all of us are excited about the coalition's launch bay with that kind of opportunity. Hope What vision of the future you leave us for a multilingual internet?

    Several things. First of all, I hope the automatic translation will be an avenue towards helping people communicate who otherwise could not. But there's another even more fun possibility. As you start interacting with these online systems, you can imagine them being tutors to help you learn languages.

    I think the

    ease with which you are interacting, it's you're literally having a conversation. These systems are not like the classic learn a language by memorization and everything else. This is an opportunity to be immersed in a language and maybe expanding vocabulary over time, you start with a modest vocabulary and get comfortable with the simple expressions, and then you get more vocabulary going on. Some of you know about a famous author who's passed away now, Dr Seuss, and you'll remember that his beginning he started with a very limited vocabulary. Remember, Sam I am. I do not like green eggs and ham. I do not like them. Sam I am. And it goes on from their very limited vocabulary. Then as time goes on, Sue's added more and more words. So I am convinced that people will be able to learn languages in a kind of immersive way, in private, you know, interacting, instead of in a big class, just interacting that language tutor. And I think that's going to make a big difference. Can I just say one other thing? You know, I'm a grandfather, and I have to talk about my grandchildren. I have two grandsons. They are seven and nine, and they are trilingual. Speak Spanish at home, speak English with their friends, and they speak Italian at school, because they're in a school called La schoola and their homework is in Italian. My grandchildren speak three languages fairly fluently, and I'm sitting here stumbling around. It's really amazing what immersion can do.

    Of advertising,

    everyone. Fince, sir, thank you very much.

    Okay,

    oh, stay here. Oh, okay, oh, the torture isn't over, all right.

    This is this the vodka that I asked for? Oh, okay, thank you.

    Hello, everyone. Super excited to follow in on that very provocative but also really interesting conversation. I'm excited to moderate this illustrious panel on AI and enhancing access to a multilingual internet. I'll begin by saying congratulations to the Coalition for launching such an important initiative and bringing us all together. I've really enjoyed just thinking about language all day, and so the question we'll grapple with right now is the one that's sort of been peppered in throughout the day about whether AI systems will break or build, break or build barriers to a multilingual, lingual, plural and multicultural internet. Before we begin. My name is Aria. I may senior policy analyst at the Center for Democracy and Technology. We are a non profit, non partisan organization dedicated to advancing civil rights and liberties in the digital age, and we are proud member of the coalition today as well. So I am here because I work on a lot of issues, including multi multilingual access to the internet. I co authored our inaugural research paper on the topic called Lost in Translation, not very original, but lost in translation, large language models in non English, content analysis applications. Two years ago, with my colleague Gabriel Nicholas, since we have published a deep dive into Maghrebi, Arabic, Kiswahili, Tamil, just last week, and Quechua is forthcoming as well, all thanks to the Internet Society Foundation, those four are low resource languages, languages with Few high quality data sets to train automated systems. And personally, I'm also very curious about this topic comes from family of translators and dubbers, including of Cartoon Network. So today we're going to be gathered to discuss, as I mentioned, how AI systems will advance multilingual access

    in the sort of

    deeper dive into the sort of applications or domains through which these systems will work. I think we've already heard a few statistics. Number one, that there are over 7000 languages spoken or written in in some sort, some form, some shape or form, but only a handful are really prominent on the internet in a meaningful way, like 16% of the world speaks English, over 60% of the internet internet's content is in English. That number changes depending on how you sort of measure content. Measurable access as well. But what little data on the internet that is languages other than English often ranges in quality. So when we add Wikipedia pages in different languages, we're thinking about community grown natively, created, high quality and often quite rigorously tested pages in non English languages, but broadly available information on the internet in non English languages are often tweets, Reddit posts that can range in quality, range in formality, range in representativeness of how people actually speak. A lot of this information is also machine translated, and we heard earlier that machine translation has often is not always representative of how people natively or the cultural connotations of language. And one big data set that's also used to train automated systems in low resource languages is often sourced from religious documents. A lot of the translations of the Bible, Jehovah's Witness documents and other religious texts often make up the corpus of what trains these systems. So what we're saying here, really, when we think about the representativeness, is that AI systems, while credibly powerful and capable in bridging the world together, translating, as Dr CERF was talking about, and allowing us to connect with one another, they may be particularly error prone or non representative of how people are speaking, and in the data sets, the gaps in sort of translation, digitized text in different languages also speaks to potentially implicit biases or prejudice we have in different languages and culture, right? If there is the very few data sets that speak to scientific inquiry in a specific language, what does that mean for our belief or connotation of whether there has been scientific inquiry in that language or culture. So the gaps that we're going to dive into, we really want to talk about the current state of what's at stake and how we can move forward. Very quick introduction, just so that we can dive into the meat fully. Jon already to introduce Dr Cerf, but this is quite difficult why he's widely known as one of the fathers of the internet. I'll just, I'll just say that he's the recipient of the Turing Award, the prize, the president Medal of Freedom, and, importantly, the most influential man in 1994 by People magazine. Next to her is Charla Gupta, Vice President and Chief program officer at Digital promise, a global nonprofit working to expand opportunity for every learner. Digital promise works with educators, researchers and tech leaders to design, innovate and scale innovations that empower all learners, no matter background. And prior to joining digital promise, Charla was the vice president of Chicago Public Education Fund. So here we're really thinking about both education and social innovation. So happy to have you, and then in the far what is this left? AJ Bedelia leads Government Affairs and Public Policy for cohere. Cohere is doing incredible work on building secure, multilingual, enterprise level AI systems and leading research and what it means to build ground up participatory AI systems and benchmarks that work in different languages and contexts. DJ himself is a trusted voice in tech policy with nearly two decades of experience in government and the private sector. Prior to cohere, he was at Google, holding several key leadership roles, advising executives and public policy strategy, and then also spent a number of years in the executive branch and on Capitol Hill, so extremely excited and hope. And I'll start with Chawla, how would and then we'll go down the panel, how would each of you describe the current state of language diversity on the journey, and where is AI helping or hurting that balance, maybe in what you've seen in your work or beyond.

    Thank you for having me here. So as I was saying, digital global nonprofit organization, we work with researchers, practitioners and tech developers to create new solutions ground everything we do in learning sciences, research. How do students learn? How should teachers teach? And then we wet those with practitioners, and then tech is more of an enhanced and we've been doing work in in education since 2017 2018 so we wrote the executive guidance that came out from the Office of Education Technology on teaching and learning in AI. We wrote guides by that office. Name is somewhere in the background there, but we are the guides for developers, for system leaders, and we are generally an AI forward organization. Said, I have my doubts. So may not have a job. Office, I hope you're hiring. But one of the things what we have seen with Ed Tech in general, is that it's not always what is, what is right, right? So learning science research tells us that the best form of instruction is mother tongue instruction, and is pushing just a few dominant languages as an example, we've been working in Haiti for about three years. My colleague, April Williamson, who's there in the back, leads that work for us, and we're bringing blended learning to very rural parts of Haiti, the central part of region. These are schools that have no electricity connectivity, and we're going in there and saying, let's do some blended learning. When we started doing that work. One of our requirements was that we should make sure that the content is in Haitian peril. Because of these past most of the content, especially in schools, is in French, and so it was really, really hard to find content in Haitian peril. We could find maybe some resources. We could find some feature resources. We could find a lot of non digital stuff. We could not find really good content in patient because the ed tech market, tech business is out there pushing things in the dominant data, rich languages, and so ended up creating our own digital versions of content we already had. But this is where we I see an inherent tension with what is possible with AI and what might actually happen with AI. You know, when you're printing textbooks, it doesn't take that much more money or effort to print a textbook in Creole versus in English. When you're scaling a tech solution, it is just easier to continue English or French rather than to create a specific solution for a very niche market thing with AI, as I think you saying, it's double edge. On one end, we're seeing AI is being used the earth species project using AI to capture and decode animal language, all its modalities, expressions, voices, everything. So you look at that and you say, Wow, AI can't do that for any more language. What can it do for our 7000 languages, half of which are at risk of extinction. You also see risks with AI, especially in education. There's a flattening of language that AI just inherently brings in, even in English, which is different. You see that AI agenda has primarily, so far, been used to streamline, to make things efficient up, to create the immersive environments that went to work. Just save for teachers time and then add more workload to them, versus saying, how do you just completely change what education feels is experienced by standard? So I am cautiously optimistic. I think there's a lot of intentionality and design that has to go in if AI is going to really expand and use of language, particularly in education, which is where a lot of students, you know, that's their oftentimes, their only exposure to the internet is, is through education. But haven't seen enough signals yet that make me think that that's where the industry is headed.

    You get another bite at the table. First of all, something that that you said, the research idea that this may be old, and Google has already done it. I don't know. I'm thinking about language training for these models. Historically, a lot of it has been text based, or it's been audio, but I don't know whether we've ever tried to combine audio and text as a way of reinforcing training for the llms in particular, for interaction and then maybe also for translation. So I'm going to go talk to my Google DeepMind folks about that, also thinking about other languages that are not audio. How about sign language? Something that's very hard. It involves whole body interpretation, not just moving fingers, but that's something that has not been tackled very effectively yet. I have an anecdote to share with you about the facility of automatic translation. I'm in Germany, Heidelberg, and I get up in the morning and I want to check the weather, because I type in weather, you know, normal, the system knows that I'm in Germany because it knows what my IP address is, and it maps it into Germany. So it goes to a German weather website. I've got chrome set up so that it automatically translates anything that isn't English into English. So I don't actually notice that I've gone with the German website because it happened so fast that I didn't see him. It just it literally goes from the German to the English, bam. So I'm reading the weather report. Does probability of fog, zero, probability of rain, zero. Probability of ice cream, zero. And I'm looking at I take my laptop over to my friend, did you have ice cream storms? Here? It turns out that the word for hail is ice and EIS, that's also the word for ice cream. And it's pretty clear that there were more instances of ice in the context of ice cream than there were ice in the context of kale. So the translation preferred was ice cream. So I sent a note to our language people saying, you might want to look into this. So you know there are ices where the training material has an impact on quality. When you mentioned Haiti in particular, I was thinking, wow, did they speak there? Because my name from Montreal, and I can tell you that, although I'm not, I have no French facility at all. Embarrassing. The French that's spoken in Montreal is not the French that spoken in Paris. And they will tell you that if you're in Paris, they will say, Well, you're

    a colonial.

    So what about Haiti? Is that a peculiar patron? Yeah, it's

    a different, different kind of French. Is that correct? Yeah.

    So it's a little bit like what happens when you get to New Orleans and you get a very peculiar to so thank you for that. That means that the training, if you're using AI at all, it's going to have to be very specialized. If you're using it, especially for for education. There's something else that's very important about the audio capability of these online systems, and that's to help people who are not literate get access to the internet. That was a observation for a long time. Many years ago, this was understood to be a potential utility. But there's something else that I had not fully appreciated until recently, that audio interaction with the world wide web may help someone who is blind get access to web based content instead of using a screen reader. If any of you have ever used screen reader or heard a screen reader? It is trying to walk someone through a two dimensional space away. It's very awkward and and because it takes a long time, blind users speed up the audio. I was I was having a conversation with someone, and we were having racing each other to see who could get an answer on Google first. Listening to this guy picking on his keys, around through the through the web page, and he's clipping it, you know, he knows what's coming next already, and so he's got it already going at two and a half times normalcy, and he's only bringing first soul bowl or two. I finally just stopped listening to this thing. Well, it turns out that if you can have a conversation with someone about the web page content, about what it's supposed to be doing for you, conveying information, doing an action that might be far more enabling, and trying to figure out how to use the buttons and where to click and audio may turn out to be hugely helpful mode of operation. Mentioned, oh yeah, there's this great multilingual joke. You've all heard this before. English is the lingua franca of the internet. Triple language pun mentioned, Earth species, I can't help but draw attention to something else called the interspecies internet. You look at Inter species internet, io, you will discover that I'm the chairman of that group, and we are actively involved in working on digitizing whale song, digitizing other sources, Earth species collaborator. We're all very excited about possibilities there. So I am, I think we are still in the early stages of exploring the capacity of these systems to engage in multilingual environments, and so I'm hoping that you will persist in what you're doing, because every single step to take the envelope of what we can make the system

    do and end of pitch. So

    just want to say I'm gonna throw for a second, which is being on a panel with Vince. He's the reason I wanted to work at Google. And is, is just an incredible honor.

    But so things,

    I'm going to play the AI hype man for a second, and then I'll give you the sobering reality of what that means in a multilingual context. So I think I'm sure you folks have seen in the news How is going to become the operating system for the world, right? It's not just us with chat bot. It's going to be interspersed in your daily life, on everything. It's going to be that platform that is the connector between you and the rest of the world, whether it's as an assistant, it's as kind of being a question answer engine I'm working with you now. That's great, right? It'll make us more productive. It'll make us better do on a daily basis. But I think what's been really well described so far has been, what does that mean if AI is only talking in English? And Alia had some great statistics about the monolingual type of the internet just generally. But another great statistic I love to use, as well as the next five languages are about 20, 25% so we're really talking, you know, 6994 languages I take up another about maybe 10% of the internet altogether. And a good amount of those are low resource languages under 1% well under 1% and a big part of the work that COVID is, frankly, Ali mentioned, we do privacy, we do security, but we spend a lot of time and research on multilingual AI. And the major reason for that, partially, is because of these indigenous and I kind of look at AI, and the challenges with AI in three buckets. First is on the training and data. The second is on the evaluation, and then the third is on, kind of the future of AI and things that multimodal context, as Matt was talking about, and also on reasoning. And so I'll just kind of walk through very quickly, just to give you a sense of the challenge we all face. So on the training side, it's not surprising that I think that to train these large language models, you need a lot data and information. And so we actually conducted, probably, I think, the largest of design, a data training project where we worked with 3000 researchers across the world to essentially train and translate a ton of information to actually create a model that does 101 languages. And the reason for this is exactly what was described earlier, which is, if you are talking to a model and it doesn't represent cultural and linguistic values, you can use it or leave even further behind. And so this additional work that's been done here. I mean, we did this as an open source project in this space, and we made the model itself, the data that came along with inclusion, display information, completely open source. And I think there's a lot of work that needs to be done we can get in the panel on the evaluation side, how we evaluate models is also extremely poor. Most of the evaluation subsets are translated from English, usually machine translated from English. So not only are we not really building these models, or we're building them poorly, sometimes we're also evaluating them in an English monolingual context as well. The third and final sleeve is as we move into additional subsets of of of of AI models. Reasoning is a really good example. As you how many of you guys have used reasoning models here before? How many of you have seen kind of what the model talks to and reasons to you back in English? And so if you think about if you were asking even a multilingual model, how you're going to a specific question, it's still going to translate that reason, it's English and come back to you again in your language. So it's translating it twice. Think about what that means for a multilingual context in terms of cultural and multilingual sensitivities that come along with that as well. And so there's a whole, I think, massive section of AI that folks don't fully appreciate. Everyone focuses

    just on the and we are on the cusp of using multi,

    you know, multi modal, a host of other things that I think is just going to make the gap even wider. And I think this is a extremely important effort of what we're doing, because of the fact that we don't start today, as we've seen with the internet, AI is going to become mind blowing as well.

    Think, yeah, we can put on like questioning perhaps on the screen, is that possible?

    No, oh, I see. That's possible. That would be

    just want to add on to just, I'm so glad you brought up evaluation and benchmarks for measuring a lot of these models. One of the biggest benchmarks available to test system performance and capabilities is the MML you ask language understanding model, 16,000 questions across domains, seen 1000 questions, domains, including math, science and but I think what we're talking about, you know, AJ was talking about, when we translate these benchmarks, yes, they may be able to ask a model or assess a model, ability to serve questions in math or Science or law in different languages, but a remit of questions, the the world view is still lies at the English or the Western level. So potentially, your models ability to answer a question about the First Amendment, but not constitutional charter in a different context, right? And what might that be a proxy for or leaving behind in terms of assessing other capabilities. Okay, I'm going to jump to charaa to ask to first plug an incredible guide that digital promise out about sort of the opportunities and limits of using chat, GPT, but also other automated

    bots. To equip

    learners with multilingual writing skills, equip learners to sort of improve their performance writing and I think one thing that I was really struck by is that you do enumerate a lot of benefits, we also talk about potential shortcomings and gaps where these systems might misallocate or mischaracterize those who speak English language as not as, you know, as performing not as well as their counterparts. And so really want to invite you to talk a little bit about that work, but also sort of what's at stake here, right? Like, who is going to be most affected?

    There we go. Oh, there we go.

    Technology. Excellent, doesn't Yeah, so we there's a lot of examples broadly, you know, people have started sharing those out in terms of where AI tools are not working. There was some study that was done that showed that AI bots are evaluating essays written by non English speakers versus English speakers, they have a much lower accuracy rate. So like 60% of those essays were by non English speakers were flagged as written by AI, that these were either plagiarized or they were written by AI. And those are some real consequences. I read another anecdote where, I think chat GPD was asked to narrate The Hate U GIF novel an American protagonist, it's Nate a voice randomly just kept throwing you full text at random points. And so there's, there's a natural consequences to sort of starting with the English AI tools, and then trying to add them to different languages. We did a research project, I think that's the one Alia is referring to, where we worked with a doctoral student who was lingual, had her write her dissertation using AI as a collaborative tool. And I think initially she thought, wow, this is really helpful. It's getting on this language I can use. And then she started sort of seeing that there were a lot of gaps that language wasn't as sharp. There was a flattening of the language that earlier. So she started using AI more as a thought partner as a generator, rather than as a movie. So it would take, she would take language and use it with her dissertation, group her advisors, and be able to refine it. And that's a great example, right? You're like, what this is, what AI could do that is not, that is not something we're all yet literate in. You review something that has been written by a human, mostly sort of reviewing. To say, everything here that is completely off doesn't read, right? We're not often always looking at it. Anything these ideas sound when you create a generated piece of document and you're reviewing it, don't always bring that lens to say, is this idea actually sound? Whatever they've collected from different parts of the internet created the lowest common denominator, then put in front of you? Do you review that? So that AI literacy piece isn't something we all have right now, and so even tools might be getting better. I don't think the we are all trained. You know, I think there is something that the idea of a certificate, of a driving certificate, or AI that is founded in AI literacy, something that all students should have. We have started doing some work here. So one of our projects right now Amira learning, which is a voice based tool, so it listens to students reading, and it gives you real time feedback. If you're struggling on a word will tell you, hey, this how you know this is what this word meaning. Do you want the meaning of this? Or where are you struggling and almost having a human being with you to lots of lots of adoption across schools in the US out. It doesn't work very well for English language because it is not used to that intonation, that voice, the different language. And so we're working with it to see if it can be trained. I'd like to see not sort of solutions that started with English and to see, hey, can we adapt them? But the solutions that start languages other than English or the other sort of dominant languages, and I don't think we have seen that yet. I don't think the research funding environment is going toward English language learners. So that's, that's what sort of keeps me, you know, gives me some hope, but also keeps me pretty realistic about Yeah,

    thank you so much. And yeah, I think, like, the most fun part of doing language and AI related research is just meeting so many, like, enthusiasts of the world who are sort of trying to, you know, identifying similar gaps as you are, but also trying to move the needle. And I'll just plug a couple, because I think they're doing incredible work. And also have, like, fun, funny names. One of them is sea lion, which is building automated systems across Southeast Asian languages. And again, going to your point of you know, there are so many communication opportunities between East Asian languages and another Southeast Asian language. Most you know. A lot of people who've like been raised outside us know that often our families speak like multi languages, multiple languages, multiple families of languages, and actually sometimes speak in multiple languages, same sentence like code mix within the language. So what are the potential for creating data sets that look like that, and so to sort of build upon that, but also talk about automated systems in a non education or non chat application. I'm going to jump over to AJ and then to you Dr CERF as well. Why is it important to offer these high quality, multilingual enterprise systems, where are people encountering, encountering them? And yeah, a little bit more about how you've been working with sort of

    experts around language.

    So at the

    end of the day, search I described, of course, is from our nonprofit research lab, and it's part of John product in terms of advancing science and multilingual and AI, but all of that work, we get completely open source. We also integrate it into our enterprise product that we set companies across the world. And I think that's a good place to start, which is, if you are a multi company, you have teams across the world, and having a chat bot or having an LLM, more importantly, that is proficient in those languages is good, but having a chat bot that is not only proficient, but proficient and speaks the language of business is even more important, right? A poorly performing think about a quarterly version of these llms in English. That's what we're dealing with right now, and even the mid resonance languages right now. Before we even get into low resource we actually three projects that we have done across the world on a business level. We have created a Japanese language LLM, working with a company called Fujitsu, a we're working on Korean language LLM actually with with LG and then we have an Arabic MLM as well, which is we focus for the Middle East, as you can imagine, also with a partner there. And the realization that we have worked with businesses is that there are things you use in a casual context, right? You might say, Yo, if you're talking to your friend, but that you would never use in a business context as well. And I think it's extremely important, because speaking the language of business not only advances your work in country, it will also help you and improve you, especially bringing those people up. It may not be as bad writing and you know things as well, bringing them up to kind of the same level of others that may be more sophisticated in that area, but maybe like me, who's not good at math, and so I think it's extremely important. Multilingual is important, not only because it helps bridges cultural divides, and I think will help and maybe perhaps improve some of the inequalities and inequities we have in the sense. I also think it's extremely important for business, truly, because we are quickly moving into a global world, you know, and especially for AI, we thinking countries around the world who want their own which their own cultural sensitivities included. And you know, the Kappa Q example is really good, right? I like to use a really silly example. In Canada, people spell the word defense with a C, not with an S, and you can tell immediately if somebody's Canadian. Somebody is Canadian American, just based on that very simple thing. And I mean, that's a very stupid example that kind of control sensitivity matters in business, matters in relations, and matters in the general social work and fabric that we deal with. And so I'll just leave it that I think none of this stuff is going to be done in all as this one big all inclusive things have to be these projects that focus on specific languages. And, you know, that's something we're very proud of as part of our IR language. We are a project because we translated 101 languages, maybe into machine, you know, actually something you can train a model into. But then we release that for people to make additional work and create additional models on top of that, and I think more of that is going to be

    necessary with a little resources. Thank

    you. And I think the trust piece is a really big one, right? Like, I think just having talked like casual about these projects, you know, I think people can immediately tell that the assumption or the tone or the world view of like these models were first released, or, you know, there was a belief. I met with a researcher who studies, who actually advances like research and testing Korean llms in Korean and and some of the prompts she was using were related to the Korean War, and the outcomes were incredibly legible and grammatically sound in Korean. But she could immediately tell that some of the perspectives or opinions presented in this perfectly grammatical Korean was actually the American view of the Korean War. And for, you know, that's something that perhaps I'm not well versed to assess, but some

    Jen is very

    adept at, like bird

    medicine. Yeah, I just want to add on. We haven't even talked about the difference between Latin languages, right? And as you move on to specifically the in Asia, and Southeast Asia in particular, and also between that and, you know, it's much more lower, lower resource languages, such as catchment, as you were describing.

    So I think we're, we're gonna move along a bit, because I have a couple of more questions to ask, and I think we're running out of time. But doctors you alluded to, we're at the early stages. So your view, what, what more is there to be done? What more is there to be done? What are some research

    you know, that's like asking the end of the quarter physics exam, describe the universe in 25 words or less. Give three examples, right? First of all, I have another anecdote for you, because if you go away with nothing else, at least you got some anecdotes. So recently, there's been a video floating around where an agent bot is placing a hotel reservation on behalf of a customer, and it calls up, turns out to be another agent bot on the hotel side, and then they discover the two bots, discovered their their agents, and they immediately switch into agent tees to speed up the whole transaction. And so the rest of the video is sounding like a 19th million. Okay, so that's always cool. Um, I think there is an enormous something that could be done this. AJ, this is a question for you. In my discussions with colleagues who are multilingual, especially ones who live in families where multiple languages are the norm, India would be a good example of that. They often talk to each other in a polyglot and I don't know what the hell our systems do if somebody is trying to interact in a polyglot way with three or four different languages all being words being mixed together, it must drive our bots crazy. How does that work?

    It's a really good question. I think that's a really good example, I not yet reflecting the world that it is trying. I will, will revolutionize, of course, but 100% I think we're not even quite there yet. I mean, how, as we were talking about, we don't have a good amount of low resource languages, even that are covered in AI. So that will come. And I'll just leave it with. I think evaluation is a big part of this, because you, you you don't own AI performs, do it yourself. And you're like, oh, I don't like this. But how are we creating actual benchmarks, creating actual standards for what that should look like? There it's in one very sort of language, or, as you fully put it, in multiple languages at once.

    I wonder if I could have a couple of other things. First of all, this is going to sound like the obvious, but sometimes stating the obvious is an important thing to do. As I listen to this conversation, realize that is a very important manifestation of culture like I say that sounds obvious, but but appreciating that it is emancipation of culture will react to the way in which their language is being used. Important. I spent some time in Vietnam, not during the the what we know more. I was there in 2023 in Hanoi, and I went to the museums. And of course, in the museums, they have exhibits, and they speak of the American War, the Vietnam War, and that's a manifestation of a different perspective and point of view. So that's another example of the language itself is reflective of people's mindsets. Example of these multilingual systems. One of my colleagues is former director of the folder Shakespeare Library, which is here in Washington, DC. And if you happen to be hanging around for a little while, please go. It's the most magnificent exhibit of shakespeariana and Elizabeth period, and it's recently renovated, $80 million worth of renovation, 12,000 square feet of additional space. I'm on the board that's held my Whitmore, the former director, has been training some of the bots to emulate Shakespeare characters. So he's got a dozen of these characters, and now he's letting them interact with each other, and I have to tell you, to be absolutely hilarious listening to false as you know, in his responding to various things. The thing I want to emphasize here is that the richness of the language models is we are still exploring the abilities of these systems. They're simply stunning. They produce extremely creative output, in my opinion. And although there was an earlier comment about maybe the flatness of of the output. I still am astonished at what these systems are capable of generating. You know, if you ask them to generate a business model, to do X, Y and Z, it's amazing that they produce stuff that sounds, you know, fairly impressive and even coherent. And it's embarrassing when you when you see these things generating these clever things, and you think, God forgot that, it makes you feel just slightly uneasy. The last thing I wanted to mention is a manifestation of

    cultural

    America. Leave it at that.

    Thank you so much. So I'm going to ask the final question, it up to Q and A from the audience, if we have any, the question is, sort of, who should be involved, right? Like, who needs to the table in developing and testing? And before I sort of invite everyone to answer, I'll give a couple of examples of what is currently moving the needle. So despite Catalan maybe not being the most spoken language in the world, certainly not 20, not in the top 30, I believe there's actually quite a number of high quality Catalan resources and training systems in Catalan, and that's because of the work of the ICANN so if you want to purchase a dot cat domain, you have to have your baby in Catalan. And that has resulted in a lot of cat and feline enthusiasts translating, sometimes in machine translating, their web pages in cat LAN. And that's one way the ICANN has sort of promoted and the cat proprietors of the domain have promoted cat LAN content creation. Another is Nvidia and the Swedish library have been working on digitizing the public trove of information so that NVIDIA can build models that fit different domains. And another great example that has actually been written extensively about in I believe MIT Tech reviews is the work of Te hikum media. So this is a nonprofit group of indigenous language activists and enthusiasts and MLB professionals building data to the Maori language and then licensing it to model developers when they find a sort of use case or domain in which they want an automated system to work in their language. And that has allowed both native creation of data sets that reflect the potential need, but also sort of owner and pride in potential uses of automated systems. So these are just some examples of, you know, cross sectoral partnerships, language that's moving the needle. So yeah, would love to hear, you know, maybe other examples and and just generally, you know, like, who should be involved in what multi stakeholder arenas, if any,

    should play a role.

    I'd be happy to ran out of ink to answer your suggestion, linguist, anthropologist, sociologist, neuroscientist, neuroscientists, educators, researchers of all kinds of politicians, bureaucracies, people and the general public.

    Say, you know, again with the education, and start with students. Take a digital promises, exhibit relation approach that centers them first and then says, what is it that you want done in your community, rather than innovation being done to you? And I think it's critical to ask them not just what you see happening or what are the potential what are the unintended consequences of success? So if we are successful in using AI to preserve language, what else happens? Do we lose? But do we lose in terms of the human role? Do we get less funding from linguists who are working on this because it's just cheaper to pay an AI bot? So I think really making work thinking through all of the possible consequences, and nobody can do that better than the people right at the center of it, and that's the community. So I would sort of say, we push for I'm

    not going to add much more to what was described, because I think the simple answer, everyone is extremely important. And I think obviously, for last century, almost, there's been a big effort underway just to preserve language, generally, right, especially low research languages that might already be endangered. I leave it at that. But I do think there's a second part of this, which is folks forget that training a model a lot of money and then running a model for inference is also extremely successful. Models. Is not our biggest model, it's our Arabic 7 billion billion parameter model, which is on potato, essentially. But the reason why it's important is because, as you start getting into these lower and lower resource languages, it's extremely important that we also think through what the costs of running models are, and thinking about efficiency and where they fit in as well. And so there's a huge part of this. And I think, you know, we say, Oh, well, we need researchers, but we also need to give researchers access to compute and access to visibility, and also thinking through, what is the end user and how is this end user going to

    be using? Just add one other thing. Whenever I work with my engineers, the first thing I ask them when they're trying to design it. What could possibly go wrong? And your point about unintended consequences falls into that category. That's a really important question to ask about. Any of these enterprises could possibly go wrong, and what can we do about it?

    Thank you. Do we have any questions from the audience or on the live stream? Oh,

    there's a microphone. Thank you,

    Katie, much. And is that used observation, which is nobody has been

    is fine because, you know, and there are times When that's unavoidable, but it does

    have their interest

    mean what you would think they would mean

    to respond to that observation.

    There are

    times when misunderstandings can occur for precise describing, where you use a term that you think you know what it means. And it turns out, this situation I got into in England, we were having a good engine, and the British partner on the other side, I'd like to table this. And I erupted that point saying, I come here, you know, 7000 miles, and you're telling me you want said, Yes. And said, I came here to talk about it. And he said, That's what I said, No, you didn't. And so, you know, this is your rights, big hazard. Well, he would have mostly expression of the US and the UK, two countries divided by a common Hi.

    I just wanted to ask about non state populations, populations and country where might be an entire state. Relationship with that, you know, with that country, with nation state, how do you ensure that those languages that might be, you know, actively erased at this moment, that chance or an option like, I know that there's generally, like a commercial interest in preserving certain languages, languages that have large markets where we want to tap, want to have the possibility, whatever, but you, I guess, to cultures and like places where they don't really have that necessary, that incentive,

    that out, I'll have a large scale answer for how do you, how do you just to pay attention to that. A lot of what we have done in education is student led around this. So one of our projects is students creating, you know, solutions to sustainable development goals, and quite a few of those focus on language preservation, using technology in a meaningful way to capture those to capture the stories and culture of elders, because those languages are addressed and using technology that the elders don't have access to students to bring those two together, I think finding ways to get that respect for and appreciation for culture from a very young age, something that we pushing and focusing on. But I don't know how market forces instead of say, let's pay attention to the ones where there is no commercial, there

    isn't a term, which is I have been introduced to over the last few years, digital humanities. And the whole point here is to introduce into humanities the power of digital communication, including large language models and related AI technologies. Getting people interested in this sometimes requires philanthropy, and the National Endowment of humanities has historically supported this kind of thing. Unfortunately, state of affairs here in the US, those institutions are severely rewarded, and that's not a good point.

    I'll just switch. I guess this is where open source and the efforts of open source are extremely important, even subsets of work people can build on. I think, I mean, I talked a lot about the IR project here, but it's one we're especially proud of because of that reason, right? We had 1000s of people across the world translating millions of pieces of data in languages that would otherwise, you know, would not be trained on or used for that purpose. And I think it just requires a couple of people right to have that dedication. And open source is a beautiful thing, actually, because as long as you have access to the internet, you can participate. And I know that's a challenge itself, of course, but, but, you know, I do think it's, it's important that we that, as we say, everyone needs to play a role. And I really do mean

    everyone. I just realized there's another anec. So I'm in a place called which is this site in arien Jaya of a very large gold and copper mine. Wife and I fly out to meet a village chief in an area called pyramid, and we land on this grass strip, and a woman comes out of a sort of a continuing roof shack, and she says, anybody here know anything about computers? And because she's having a problem with her, she's translating Bible into the local languages. The guys who's with us just have in his pocket. This guy has to be uber Gee, he's got in his pocket Norton Utilities just get and so he's busy transfer hard disk drive while she's making, you know, donuts for us. Here we go to visit the village chief. But the thing is that there are dedicate themselves, as you imply, to doing some of these things for the sheer way of having done it.

    I was going to add your point of open source. Respond to that. I think pushing for more, right? So Wikipedia has more language support just then, Google, Facebook, all of them, signal. The open source messaging app has again supported more languages than WhatsApp does. Shin yet, but still. So I think pushing for for more of that, pushing for more open source resources and letting users have have a role in expanding language support.

    We evaluate the quality of those open sources now we this for open source code that often doesn't get maintained very well, and other people get into, you know, very So is there an elevation that you can do?

    Question from the online audience as well? You had talked about the different stakeholders who should be and someone is interested in learning specifically about the role language experts and others in academia in particular,

    I'll just say in our research, what we found is a lot of the work is being done, both in open source and outside spaces. Increase resource of low resource and medium resource languages, or create or even test and rigorously identify gaps in current training corpus of big models is done at the academic level, and I think the biggest, one of the biggest barriers they're facing is just a access to the companies To then, like, scale or implement their findings or recommendations. And B take research and scale it into different applications or languages, and that's both a sort of access to company issue, but also funding issue. We need a virtuous cycle of research, research to begin with, research, and especially outside the United States, I'm aware, maybe in the United States, a lot of different funding sources for academic research or independent research, or other countries are finding that the sort of budget or line item that is going towards academic studies is 0.002% of the GDP designate For science, much less NLP in languages that are, again, maybe not commercially viable. So huge. Plus one for like academic focus on languages. And I'll just do one plug for the research done by Emily Bender, who is a computer science professor at the University of Washington. She's going to very, very helpful sort of nose rule called the bender rule, which is a call for NLP professionals or academic researchers to name the language they are using, language that their research pertains to. She finds that the bulk of research and development in automated capabilities, or specifically, doesn't note the language they're talking about. Or the studies are significant or significant, and it's because the assumption is that it's on English. And so she's saying we actually can't measure asymmetry between advancements in English, because we actually are don't know. I mean, we know based on our own observations, but we don't know. Anyone else want to know. I saw a bunch of hands,

    yeah, just coming from the point of academia. So I've been working on one when talking about ice and ice cream. And I was fortunate enough to be one of the first projects that worked with open AI at ASU, on on using it on the academic side, testing it, piloting students. And, yes, great excitement, also some disappointments. But what I'm seeing now, so German, for me, is basically only like a prototype, like all like a use case, because it's one of the languages that has a large amount of materials online, so it should work as one of the best, right? But I think what now we're at the point where we have to look across the languages. I'm sure there are certain things we're very often working in our little silos, as much as we try to work together to see what are things that are universal among languages, that might be misunderstood, and what are particular based on the language or in which way you go? Well, yeah, anything. A lot of potential, as long as we start now going out of our little corners and really see it more as a universal effort.

    Thank you very much for that question. Actually, with the Unicode Consortium, which non profit, open source, open standard body for internationalization, we are at the foundation for how people can communicate in their languages online, and we're always working with academia and linguists along with engineers and program managers. So thank you.

    No, I love these sessions because it keeps giving me ideas for research. Imagine for just a second that we assemble this image of language, grammar and sound, and then we start to ask questions among languages. Think about the Romance languages we know very clear relationships. Even have some idea about the historical evolution of language, going from old German or Old English to modern and modern English. You can see various place. What I'm saying out here is is trying to work your way backwards. How did ancient language sound? We conceivably use the knowledge we have of the current spoken languages try to predict what they might have sounded like as we work in time to the root languages. I mean, what did ancient Greek actually sound like? We know what modern Greek is, but it's not clear what ancient Greek actually sound like. And to give you a really weird example, in the rural parts of Virginia, if you go there and you talk to people, it's easy to understand them. And I've been told that what I'm hearing is a variation of Elizabethan English, very this is not the Oxfordian BBC that you expect from Shakespeare's plays 17th Century and 16th century Shakespearean things were rendered in Elizabethan English. I don't know if you've ever gone to a website and tried to listen to a Shakespeare play. It's hard enough to listen to it is Elizabethan, ancient talks, 49 accent. Listening to it in what is purported to be the Elizabethan period is even more weird, it would be fascinating if we could work our way backwards there. Thank

    you so much, and thank you to everyone on the panel, but including Dr SIR for giving us so many ideas and potential research topics. So we'll definitely be writing those down. Thank you to the panelists for a really, really fascinating conversation and all the work you're doing in advancing multilingual access technologies and thank you to

    I will let Sarah.

    Thank you so much.

    Recording stopped,

    so much so we are at the recording in progress. We're the culmination of today's event, and it's been just joyful to listen to not just the ideas about what we do language, but also about where the future, and that's a big reason why Cody is here. That's why the coalition is being being created. So I wanted to Teresa swinhardt. She's the SVP of global domains and strategy at ICANN, where she oversees Strategic Initiatives implementation, as well as stakeholder engagement trees. And I go back a long ways, and she's got just extensive experience in governance, multi stakeholder collaboration, and she's been a key figure in shaping global domain policies, but is also playing a key role in shaping, helping shape where this coalition on digital impact might go. So with that, Theresa,

    where do you want me stick there? Okay, I'll sit here.

    So there we go. So this has been an amazing session for multiple reasons, and I think this last one culminated in many ways of the multifaceted aspect of actually being able to engage online in the language and structure of your choice. And we heard about the use of AI, we heard about the use of Internet addresses. We heard about languages that are dying from Rebecca what's happening in that space? I think, holistic standpoint, we're at a culmination of being able to maybe turn the tide and bring the experience of the offline world to the online world in a way that maybe makes a difference in lives, in their cultures. The thought of cultures and language disappearing is heart wrenching. It takes away from one's family, it takes away from one's values. It takes away from one's culture and from one's identity, and I think that's not to be missing the work that we're doing and trying to keep that ahead of us. But that doesn't mean it's easy, the accessibility, but the tasks ahead are much harder, and the ability to get technology streamlined to allow for that seamless interface, was experience still is a challenge. I think, as we heard earlier today, the internet started in English, in Roman character sets, I get that trend to send itself in a different way. I had the internet started in a different part of the world, it would have been a different thing, and we'd still be facing exactly the same challenge other language groups on online not lost on the international community either. In the internet governance space, we've heard through the World Summit on information society, the Akron soup declaration that recognized education knowledge and information and communication at the core of human progress endeavor and well being, led to the commitment for a multilingual, multilingualization of the internet as part so even from a governmental space. And governmental space that's important we're seeing UNESCO take on initiatives, many of them, including a decade for indigenous languages. And those are important initiatives and shouldn't be on us in the work we're looking at. We of course, hear about the economic and social benefits that this brings. I would really emphasize the social benefits, economic benefits are hot, but that's the benefit of just being online, regardless of the medium that man. And so we heard today also about the Domain Name System, the importance of universal acceptance, and this initiative is an important endeavor around that. And we heard about the partnerships that are important to do this, because no single engine can can do it alone, and I think that has to be reinforced where we partner with each other, and what can we do? So one might ask, what ICANN is doing more specifically, in this area, we have a limited remit. There's a limited area of what we can do. What we can do, though, is our area, and then partner with institutions and organizations where we can, we have a universal acceptance and international team that is very focused on this, I think, be listening online. I certainly have seen making sure that I'm aware of the discussions that are happening. And it wasn't a long ago that when the last round of the new top level domains were launched, that there was a realization that universal acceptance of those was one of the challenges being faced that is more than two or three letters to the to the left of the.to

    speak. I think you remember those Well exactly.

    And so there was the formation of the universal acceptance steering group in 2015 to try to address these gaps we had been supporting, and we have supported that initiative, and they recently also concluded a decade of really impressive work around this work in that era, our strategic plan continues to emphasize the importance of it and looking into the future, we have just announced a plan to form a new universal acceptance expert working group to really focus in on producing specific guidelines For the organization of how to best directly implement universally, since we're now focused on the implementation side of the work, and that includes reaching out to different organizations, open source software and institutions, to read the awareness around that I was recently at an event, was example of some of the partnerships that we do in

    we have a

    supported an effort around a two day on which was really fun and interesting. I've never seen so many students inspired creative and using artificial intelligence for problem solving. And on the event that was co sponsored together with the telecommunication Regulatory Authority and the American University of Bahrain. So that was a really fun event that we did there. But we're also undertaking initiatives on education and building in the curriculum into university programs around that, by example, with the American University in looking at the curriculum, and then looking at integrating that kind of curriculum into universities in Africa and Latin American regions as well. So that's part of really getting the next generation, those who are being software developers, those who are being trained to understand that the ability to have universal acceptance, whether that is the integration of top level domains or other tools into systems, or the ability to use your language, is is actually feasible to do. And how does one build that mindset into everything on the operational side, I think we've already heard that we're working on various including the development of rules to allow for specific scripts, the formation of generation language panels that are focused in currently, 26 unique in the zone, working on more. We need the community to work with us in order to identify how to get online and support those right now, there's more than 150 internationalized top level domains, including both generic and country top level domains, that have been delegated as we go to open the Next round in 2026 that is the opportunity to apply for telemedicine. We're also very hopeful that local communities, new language, the ability to do that, are going to be part of that application in order to really bring the internet to the global community overall.

    So with that, that's the

    organization side, then the side reaching out to partnerships such as these that are here universal acceptance. Say, this is the third year of the event, and we see the participation increasing and the awareness increasing, the different kinds of approaches to the dialogs, and the amount of local and regional initiatives that are showing up around it. We also have an opportunity here to co sponsor a UA Day event in Vietnam with UNESCO. So that was a first time that we had that opportunity, and look forward to more around those so as I listened today, and we started off with preparing languages, and we end up in talking about artificial intelligence, I'm really struck that maybe we are at a pivot point and creating more awareness for everybody. So thank you very much for having me.

    Thank you for the dialog that you

    a few closing comments and and thank you all for being here, little summary of what's what's, little summary of what's to come with with Cody. So this is the end of universal Inception day has doubled as the launch of

    on digital impact.

    While we were preparing for this event, last night, I read a chapter of a book I've been reading. It's it's a really book called The Gutenberg parentheses really interesting book. It's about the printing press, but with with insights as to what we can take that when it comes to the internet, finished the chapter to finish the book on when they started working on 200 200 years after the invention of movable type. So in some ways, we are making pretty progress, right in comparison. But you know, as we heard, with all the advancing technologies there is, the world is moving even faster. So it's extremely timely that the work that we've been doing for a long time move really quickly. You and I, Ram were sitting like this, probably almost little over 10 years ago, helping to start another program called universal acceptance spring group, the kind of important work that Teresa was talking about, and the kind of organization had did, and the kind of work that ICANN continues to is vital to advancing these important technological next steps. As we close, the one thing I wanted to say and why I'm very proud to be standing here with rum to start this new group coding. It's chose a coalition start next, because it's not one group that needs to be working on this is all different groups. Had a list of all the different groups, coordination of all the different organizations that need to be working and pushing this forward is the most important thing that we need to do next, and that's vital effort that we're going to try and sort of take on, is we're very excited to have you here at the beginning.

    Yeah, and Christian, I think there is a urgency, right? We've spoken about of that. You know, it began with a conversation Rebecca, you and I had, and it's the urgency is driven in my mind by the fact that we are, we may be facing sure or issues like that. The urgency, I think, actually, is that by doing the same things, you can advance, you can have a real impact. And in our small lives on this earth, if we can come together, build a coalition that can advance understanding, can build some education awareness, perform some research that actually allows us, all of us as humans, to communicate better and to leverage the new new technology, the internet right up To be able to come together, there is a greater purpose that will hopefully outlive us. To me, that is a foundational component, and that's why, when we conceived of this idea, we didn't think it should be housed inside of an institution. You know, when we started up the universal acceptance tiering group. Let's do this inside of ICANN, great deal of relevant significance inside the domain name system. We're talking about something far more than add here. We're talking about human connection, and future human connection that comes with language that's at the foundation of all of that, and it's overlaid with technology, overlaid with tremendous advancements that are in front of us, and think is all of institutions here, and the institutions who are not yet aware of what we want to do, but whom, hopefully you will test and say you should be connected to these folks. You should be talking to these other folks, right? That's That's ask, not only join Cody, not only become a member of Cody, but bring your intellectual belief to bear money is important, we'll have to go and find funding and raise the money that's always an important thing, but in some ways, perhaps, is lend us your intellect and share your network and tell us who else we need to go interact with. More focused on this year, starting up a couple of small projects. I was speaking with a few folks saying, listen, these ideas are great. Show something, even if it's small. You know, small successes breed greater successes, right? So that's what do you have ideas on projects that will be relevant and that are, you know, that will be important to move the ball a little bit forward

    years

    about that, you know, and really need your support, not just at events like these, but the real work now begins right this. This enterprise is now taking off. How come the hard work, the real work, for what else has to be done? Support. Your help and please join. And for those of you who do you're gonna you're hearing about some projects that we were having seven months we're very excited to have you. I'm, I'm, I'm inspired by a couple of different things. For all the fears that we sort of talked about could come of inaction, the fact that, remember, back at the beginning, you talked about a lot of youth that were interested in having a lot of that they have agency around, going on there and trying to fight language preservation. My My daughter is here. She's interested in this. It feels like a movement, and it's a movement that we can help to coordinate. I'm very, very thankful that all of you are here today wondering if our colleagues over here have a couple of Yes,

    I never pass up an open microphone there from Shakespeare, specifically Julius Caesar, which you'll be familiar with, which is perfectly fitting there. There is a tide in the affairs of men which taken of the flood he's committed all the voyage of their life is bound in the shallows and in miseries. So let's grab the photos.

    Perhaps our event, there are still refreshments here. Feel free to indulge. And if you'd like to get out on the deck and enjoy some sun, do that as well. Thank you so much for coming. Thanks for everybody online. We are done

    recording stopped. You.