Select Indigenous Knowledge and AI

4:23PM Jun 9, 2020

Speakers:

Ahmed Riad

Keywords:

ai

language

duolingo

learn

stories

culture

food

indigenous knowledge

cherokee

great

sustainable development goals

create

tracy

part

recipes

teach

alva

questions

victor

share

Broadcast

morning, afternoon and evening, welcome to the AI and indigenous knowledge solution track webinar. We hope you have been healthy and safe, your families, your friends, my name is Emma geode from the IoT with International Telecommunication Union, and I have the privilege of introducing today's webinar. International Telecommunication Union is the United Nation agency for ICT. And we are also the organizers on AI for Good Global Summit alongside the XPrize Foundation, an important ship with 36 United Nations agencies, ACM and Switzerland. The goal of the summit is to identify practical applications of artificial intelligence that can advance the Sustainable Development Goals. And we try to scale them for a global impact. Now like most of the world has conducted, they AI Summit also has gone digital We're working on weekly elements webinars, solution tracks that will allow us to reach even more people through our 2020. Today's webinar is part one of the AI to preserve comprehensive solution track that would have been taking place in Geneva where it's not for the buyers. Today, we'll start with a keynote presentation, followed by a panel discussion and demo video in the lightning talk. You can certainly ask questions throughout the webinar using the q&a feature. The panel discussion the lightning talk would be facilitated by the var, or the lamb founder and CEO of Ivonne and she's also the co chair of the AI to observe culture heritage solution track. She will be introducing the speakers on the panel later in this webinar. Before we start with our keynote opening presentation, I would like to ask all attendees to use the chat functionality and let us know where they're connected from like, which country which city just have an idea of who's in the room. So I would really like to see wherever you want disconnected from how we have Geneva, Lebanon, Washington. Thanks for California, random. Brussels, Cairo, London. Wow. Everywhere. San Francisco, Tokyo, Nagano, Jamaica, Cote d'Ivoire, Poland. Okay, that's that's a great color from everywhere. I mean, I can see how geographically diverse this webinar is. And I think now it's time to start. The keynote opening presentation is going to be by very subtle, the head of research and he I asked you, Hey, can you please open your video mic, please.

Thank you very much.

It's a pleasure to be here. Thank you very much. Where are you connected from? I mean, where are you talking us from right now.

Good morning from Pittsburgh, Pennsylvania in the United States.

Nice and what time is it now?

It is

10am Okay, so it's a good time and how does tuition there? I mean, you still work from home or?

Yes, very much. So they the Duolingo. office, I think March 16, is when we started all working remotely. And that has been the status quo for several weeks now. And for several months to come, probably.

Yeah, we understand kind of kind of the same thing here in Geneva. Okay, I think you can just get ready and try to share your screen with the presentation. And while we're doing this, I would like to remind all the attendees to be asking question we're looking, we're counting on you for a very interactive discussion throughout the presentations, and we will be reading the q&a, the questions in the q&a, and we're going to have a q&a segment after a various presentation for him. And of course, we're gonna have another q&a segment after the panel discussion with the board later on. So we're pleased this was all yours.

All right. Thank you very much. I Now that I'm in full screen and sharing mode, I can somebody give me some signal that

everybody can see my screen.

Everything works really well. And you can see it's

great. Thank you. And I can't see the chat. So if people pop in with questions, could one of the moderators just pause me or something?

Yeah, sure some of the chat in the queue, and we'll be looking into this and after.

Excellent, thank you. Well, again, good morning from Pittsburgh, Pennsylvania. It's It's a pleasure to be with everybody. And I'm excited to talk about how we're using AI at Duolingo to teach the world languages and particularly, to preserve languages through teaching various indigenous languages. We have had several efforts over the last few years that I'll be talking about. But first of all, since I'm coming to you from Pittsburgh, the city of bridges There are more than 400 bridges in the city of Pittsburgh.

Here's a beautiful picture of

Actually, I'm not sure which one this is because there are three identical bridges, the Warhol, Carson and Clemente bridges. But the funny thing is, I want to convince you that language shapes the mind the way we think about the world. And one good scientific example of this is a study that was done in 2002. Showing that if you ask native German speakers are well, first of all, consider that the word for bridge in German, the Buddha is a feminine word, whereas in Spanish, it's a point A which is a masculine word. And if you ask native German or Spanish speakers to describe a bridge to come up with adjectives for a bridge, the German speakers will come up with things like beautiful, elegant and peaceful things that are stereotypically feminine pronouns whereas Spanish speakers will say strong and sturdy and towering. And just before you think that maybe there's something going on with

the, the just maybe

Spanish speakers are more macho and come up with more masculine sounding words you get the opposite effect for other words like the word for key. German speakers would say hard and heavy and jagged, whereas Spanish speakers would say golden or intricate and lovely. So the way words, things,

linguistic concepts, such as grammatical gender

influenced the way that we think about these objects, and it influences the way we interpret the world. So here's a completely different example. The gooey is mere language in Australia, and indigenous language that the people there when you talk about direction cardinality. I would say, I assume everybody can still see me but this is my right shoulder. And if I turn this way, it's still my right shoulder. And if I turn this way, it's still my right shoulder. But in this particular region of Australia, this would be my, let me get my bearings. This is my North shoulder.

And if I turn this way,

now it's my West shoulder, and my East shoulder. The the description of my body parts or any object direction isn't tied to me, as it is, in most western languages, it's tied to the land, the landscape itself. And so cardinal direction is so baked into the culture and therefore the mindsets. There have been studies to show that these people more so than speakers of other languages just have an intuitive compass, a sense of direction.

In fact, I think in this language,

if you say,

Hello, you you literally ask which direction are you going and then the answer is something like Like north northwest, how about you? So you literally can't get past Hello without knowing which direction you're pointing. In addition, there's evidence that bilingualism or studying other languages can improve cognitive function and can delay diseases like Alzheimer's. And also the learning multiple languages is associated with greater empathy and social awareness. So these are just a few of the reasons that we created Duolingo A few years ago, the fact that the languages shape your worldview and can be beneficial both for you as individuals and society as a whole. So we started Duolingo in 2012. It was a Carnegie Mellon research project here in Pittsburgh. Today there are more than 300 millions learners across the world. learning languages we have 90 courses, including the teaching Irish Which is an indigenous language, or Esperanto, which is a constructed language. And we're expanding before the end of the year, we hope to expand to more than 100 courses including Maori in Yiddish, and all of the learning content is free. And this is the part where I would normally pause and ask the audience who has heard of Duolingo, or who has used Duolingo with a show of hands. can't really do that right? very easily. But I'll give you a little crash course in it just in case you're not familiar with the lingo. So it is a an app and a website that is a game like curriculum for learning languages. So we have these lessons that are grouped into skills by topics so those topics can be thematic, like phrases, or interest or phrases or travel or restaurants. Or they could be also grammatical topics to teach teach you the case system. Have a particular language, etc. and learners complete these skills to unlock more rows. So this is how we implement mastery learning. And in order to progress through the course, you have to complete all of the skills above it. And then we also encourage users to practice what's what they've already learned to go back. And I'll talk about some of the technology we use to improve learning through that. And then there are other gamified motivators and metrics like in app currencies and leaderboards, and things like that. So brief word about AI and Duolingo. In general, when you talk about AI going into any kind of field, and people get a little shaky, and our goal here is obviously not to replace great teachers The best way to learn a language because language also is so intertwined with culture. The best way to learn a language is to talk with people and to be immersed in it until To learn the words and the customs and the traditions of the people who speak that language, but until everybody in the world has equal access to great teachers, which is kind of a stretch, we believe that AI is the best way to scale this one on one private tutor experience to everybody.

And when you think about

why a private language tutor or a group of dedicated people to practice with why they're so effective, I argue that they have three properties. One is that they know the material really well. They know the language. They know the ins and outs of it, the nuances of it and the culture. They know how to keep you engaged and excited about the material. And perhaps most importantly, they know how to get inside your head because you have this dedicated face to face time with them. They they can get an intuitive sense of what you know what you don't know what you struggle with. And then use that to their advantage to engage you with the right material at the right time. So dive a little deeper into how we use this into the lingo. So we have a statistical AI model that learns what learners know and how well they know it. Since we have 300 million people doing half a billion lessons every single day. We've got a giant database of individuals interactions with different linguistic concepts like words and vocabulary. And this is personalized for each student and adapts over time as they learn more. So how this works harkens back to something called the spacing effect, which is more than a century old. And it's the idea that people learn better if practices spaced over long intervals instead of cramming you you practice. You come back to things periodically to practice. And this is one implementation of this is the forgetting curve. So here the Delta is the amount of time it's been since you've last practice to word or concept. And if there's some notion of a half life, which is ah, let's say it's one day, then if, if the delta is zero, if I just learned this, or I just practiced it, very high probability that I will will remember it if you quizzed me on it just now. But if this half life is a day, then after a day has passed, there's a 5050 chance I've forgotten it. And then after two days, 75% chance I've forgotten that then the probability keeps cutting in half. And then there's a related concept called the lag effect, which is that people learn better if the spacing gradually increases over time, or in other words, the half life increases with more practice. And the way we implemented this into a lingo at the beginning, was this a flashcard system that went back to the 1970s where you had a box, the one day box, the two day box, the four day box, and if you got it right, you moved it from the one day box to the two day box and the next time to the fourth day. But if you got it wrong, you you demoed it. And then we noticed that this is actually can be formalized as an equation of two to the number of times you've gotten it right, minus the number of times you've gotten it wrong. which you can then generalize into to to the power of a dot product of any arbitrary thing you can think of what is the word who is the user? You know Is this a noun is that a verb is an adjective Is it an interjection and then learn weights empirically, from the record of millions of users getting billions of exercises correct and incorrect. So here's a cartoon of what's going on with a student, their first six encounters with a word on Duolingo over 30 days. The vertical axis is the proportion of time they got that word correct in an exercise at that particular time, so each x represents a time point.

It's what I said. That is the recall rate

Then we also have the lag time we know how long it's been since they last practiced it. And then we had kind of have these feature vectors like the number of times they've gotten incorrect or incorrect in the past, what is the word etc. And all of these go into what we call half life regression, which is a novel AI algorithm that we proposed in 2016. And this will make predictions indicated by the red dotted line here and we, we can fit this model so that the predictions go through those X's as closely as possible. We collected a bunch of log data and our methods using this machine learning this this half life regression specific, memory focused machine learning method was significantly better than other existing schedules, Spaced Repetition heuristics as well as other off the shelf AI baselines like logistic regression. And then when we put this into Practice, something that we do is we link over to everything. We run controlled experiments, so half of the users will get control, which in this case was the old flashcard algorithm. And then half of the users will get the new AI model under the hood. And then in this case, we ran this experiment for two weeks, there were 3 million students. And we saw a 12% boost in engagement, people continuing to come back and learn languages the next day after having used the app the day before. So some of the impacts of this, this is a timeline for the first, I don't know maybe four years or so since we launched. We launched almost exactly eight years ago, today. And when we put these models into deployment, I would like to think that has something to do with the increased growth we've seen over the years making to the lingo, basically the number one language learning platform online. We've all released several peer reviewed research articles as well as datasets. The dataset that we use to fit the model I just described is publicly available and lots of other researchers have been building on that work and continuing

to publish on it.

And there's also an interesting dynamic with the way that Duolingo is structured where users have a choice in that they could choose to plow through the course. Let me back up for a second, because I'm not sure I explained this earlier. When you do a skill. Initially, all of the content is at an easy sort of level. And then when you go back and practice it, it gradually gets harder and harder. So this is again part of the adaptive algorithms. And in fact, the reason my face looks like this right now, as we add a new model we started building in October, I started growing a beard. When that started. We launched that model a few months ago, I started to shave it off. Everybody said, No stop there. Looks great. Just keep the so these are my quarantine lamb chops.

Anyway,

so every time you go back and review the material, the little numbers in the crowns in this image here, the one is you've, you've gone through the skill one time and the two is you've gone through it the second time, which is slightly harder. And the third time and fourth time they get harder and harder. And so users have a choice whether or not the material they've already learned whether they want to keep repeating it to two, go deeper into those skills, or do they want to keep everything easy and plow straight ahead? So what we're looking at here is kind of a heat map of the distribution of what users actually do. So they can either plow straight ahead and unlock rows all at the easiest level or they can level up and there's They go through the curriculum slower, but it's more challenging material. So the people along this whole we call them the candy crushers, those are the people maybe they don't actually care about learning language very much. It's a fun distraction. It's it's a, it's like Fairtrade organic junk food. At least they're learning something while they're playing a game. But then there are all of these other people on the opposite end of the spectrum, which we assume we call them strivers because it's really important for them to engage with the material and learn the language and they're in it for the long haul. And interestingly, if you look geographically where these people are distributed, the candy crushers are mostly in developed countries. And the strivers are mostly and in developing countries, there's kind of this dynamic not entirely, but there's this dynamic that you start to see and interestingly Also, if you look at At the distribution of where our revenue comes from, through subscription services and things, again, all of the learning content is free. They're just kinds of fun game like bonuses and power ups that you can purchase and that is how we make 90% of our revenue. So essentially, we have people in developed countries using Duolingo, for fun, who are subsidizing free education for people in largely in developing countries who are learning languages to improve their socio economic situation, or reengaged with their indigenous culture. So just some examples of who is learning on Duolingo. There's a wide variety of people from all over the world. Bill Gates is learning French, the San Jose Earthquakes, soccer or football team, depending on where you're from. The manager is Argentinian and speaks Spanish the rest of the team speaks English. They're using Duolingo to learn each other's languages to function better as a team. Here are some chose schoolchildren in Indonesia. And interestingly, the number one language to learn in Sweden is Swedish. And the number one language to learn in many United States cities like Miami is English. And here's the point where I would ask the audience a question of why you think that is, since it's a little awkward, I'll just tell you it's it's largely refugee and immigrant populations. So here's an example of Knorr who's an Iraqi refugee living in Turkey who's using Duolingo to learn Turkish and German. But it's, it's not just refugees. Again, there's all kinds of people learning all kinds of languages for all kinds of reasons, including Talia here, who's a high school student learning Navajo and Blanding Utah. A portion of Which is a Navajo reservation. And so a few years ago, we started getting involved with creating courses for minority indigenous languages. And, and even today, when you ask, we have a survey of why people are learning languages on Duolingo. And most most people it's, they say it's for work or for school, or maybe they're traveling. But when you ask the learners of the indigenous language courses that we teach, they are 1.3 times more likely to say, because of family, and two and a half times more likely to say, because of culture. Now, we don't know if this is to reconnect with their own culture or to learn more about

the surrounding culture. But these are some striking statistically significant differences between the overall population of language learners. That's interesting. So we started down this journey in 20. 16 we were approached by some actually high schoolers, who were native Irish speakers who just wanted they love Duolingo they had been using it to learn French. And they, they just wanted to help us create an Irish course. So today there are more than a million active learners. And by active learners, I mean people who've done a lesson in the last week. There are a million active learners relative to 100 or 200,000 native speakers in Ireland, and you can see heat map of distribution. You know, it's mostly concentrated of people in Ireland, and then the greater English speaking diaspora, including the United States. We actually were given an award by presidents, the president of Ireland for the preservation of the Irish language. So here we've got Karen Tsai who's one of our senior engineering managers who, who traveled to Ireland to accept

the award.

We We followed that up shortly afterward with Welsh and today there are about a half a million Welsh learners, mostly concentrated in the UK. There are we also later that year worked with the Peace Corps to develop a language for a course to teach Guarani, which is a native South American language largely spoken in Bolivia, Paraguay, Argentina, and somewhat in Brazil, I believe. And so you can see sort of on the map where that's largely concentrated. And interestingly, this is most of the languages that are all of the courses that we teach on Duolingo are pairs of language so they assume somebody is coming into it speaking English, for example, and wants to learn Irish as an English speaker. Well, Ronnie is actually one of the few courses where you can't actually Learn it as an English speaker. It's a course to teach Spanish speakers scrawny. A few years later, we decided that we wanted to start the practice of releasing an endangered language, a new endangered language course on indigenous peoples day, every year. And so we are first year of doing this was in 2018, where we did a doubleheader of launching Hawaiian and Navajo. So each of these have hundreds of thousands of learners and more learners than there are native speakers of the language.

What's gonna, you know, lots of coverage and started this kind of narrative about how these online language courses can preserve the language and the culture of the people.

To which

Alistair Allen who's a member of Scottish Parliament actually petitions Duolingo to create a Scott Gaelic course which we did for our major release last year. So again, almost half a million active learners of Scottish Gaelic, mostly in the UK in the United States. And we're currently working on several more kitchen yucatec her two Mayan languages, which we're creating again for Spanish speakers to learn those languages. My Melody is under developments, Haitian Creole and Yiddish. And your definition of a, an endangered, sorry, an indigenous language can vary. There are also other languages that we teach like Swahili

in the course as well.

So here's Claire Mei. She's one of the high schoolers in Utah who helped develop the Navajo course. And so we'd not only use AI to adapt and To make teaching more efficient, we also use AI to help facilitate the development of the courses. And one way of doing that is if you answer a question that you got wrong, but you think you're correct, you can submit a report. And but only about 10% of those reports are actually wrong. So we hit about half a million of these every week. So the volunteers who create the courses and maintain them, don't really have time to look at all of them. So we use AI to rank order them essentially, like here are some prompts in Navajo. Here the suggested English translations as part of the lesson, or the number of reports and the probability that that particular translation is correct according to the our AI models, and this is what it looks like a little bit. Here's an example of a prompt. Here is a sort of, if you're, if you're familiar with regular expressions, a language of trying to capture all the different ways you could correctly translate this and then These are things that different users and submitted, we can extract features by comparing all of these different things and labels come from the decisions that the course developers made. and here we can see on the horizontal axis is the size of the course in terms of the number of reports we get. And the vertical axis is the area under the ROC curve or quality measure. So point five is essentially random guessing you're not doing anything smart at all. If you use a heuristic of just ranking of report by the number of people who submitted it, which seems reasonable you you only do a little bit better. And each dot here represents one of our courses. And even though all of these languages have very few linguistic resources to train the AI models, they're actually good enough to be useful and speeding up the development of these courses. Historically, once we created a new course it took about six months to graduate from beta and Scots Gaelic, which is one of the first courses To use this AI model, after we put it into production only took five weeks to hit that benchmark. I think I'm running a little over time. So I'll just rush through the end. I wanted to leave you with this quote from Clayton long, who was who spearheaded reached out to us about creating the Navajo course I said this collaboration was a dream come true because it reaches young people, makes them the leaders, but also exposes the world to our language and our stories in a way that I could never have accomplished. And so with that, I will

pause for questions. I have lost track of my cursor.

There we go.

Awesome, and I think you're muted.

Thank you very much for those amazing, super insightful, I think a lot of people were commenting about how interesting these slides are and you gave us but I mean As you know, we have a bit over time now. So I think we have to choose just few questions. However, you have the opportunity to answer all these questions from the q&a functionality throughout the webinar. So I think this is a very interesting question we're having here. Someone is asking, what are the downsides of using AI to teach indigenous languages? Is there any downsides to that? And if there is, where are they?

Well, it depends on your definition of a downside. One limitation of using AI is because AI to function well as really data hungry. And if if you're concerned about if you're motivated to teach indigenous endangered minority languages, by definition, there are not that many resources for them. So there are some, like I showed some promising results here where we were still able to train useful AI models. Despite the lack of lots of resources. But that's one challenge in doing it. I guess, in terms of a broader sort of societal disadvantage sense. Again, languages is all about culture. Or it's largely about culture. And so if you want to learn an indigenous language is because you want to connect with the culture or the people of that language. And using AI as a medium to learn that, I'm actually not. I mean, you'll obviously be missing out on that. But it also I think, what we've observed is, is that for beginners, especially in a language, working with, like, an app or computer system or an AI is way more comfortable than dealing you know, the performance anxiety or the embarrassment of talking with an actual person and screwing up So it actually I think helps get people up to a baseline level of fluency so that they can start to engage more comfortably and meaningfully. And, and for languages that are going extinct, this might be the only way that you can really engage with them at some point.

Okay, again, because of the time and ask you one last question, if you can keep your answer within 30 seconds would be great, because it's the one I got many likes here. So I think the question is, says my question links to the lack of enough content in non English language, if you were to bring in a new language to teach, how much do you depend on the existing internet content in that language? So how much do you depend on the on the current existing content of this language on the internet do depend on that or not?

Well, in terms of training, the AI models are so we we use things like corpus linguistics too, as a starting point for the curriculum like high school Frequency words we teach first. And then once you're if you're familiar with the Common European Framework of Reference, or the CFR, there are these six levels a one, a, two, B, one, B, two, etc. Most of our courses get you through to an A to B one level at this point. And for many, many low resource languages, that might be as far as you can get. So that is a limitation that works and sort of fighting against.

Yeah, got, you know, great, super insightful, I think, please, I mean, you can stay with us on the webinar. And can Haskell answer all these questions, if you want, and I think we're ready like 10 minutes of your time so much again, and we're going to now start with introducing continue with introducing our next segment of the webinar. So the var, the var avalanche, the founder and CEO of Ivo, I'm the co chair of the I observe culture hunter solution track Why thank you very much for being with us.

Thank you. Thank you so much. Great to be here.

So the var again for the time, I think the show is yours right now. So you started using your speakers, you start with the panel, hopefully a demo video, and also the enlightening talks, and then q&a. So please. Great.

Thank you so much. Sounds great. It's incredible to be here. I thank you, Bert, we learned so much from you looking forward to following up as well. I'm joining you from Washington DC during a very historic time, not just in our nation in the United States, but all over the world. And never been a more pertinent time to be able to talk about how we can preserve our culture, and whether artificial intelligence can be a tool in that way, as Burr has also mentioned. So our startup is called Ivo and that stands for intelligent voices of wisdom. We are an Amazon activate and new for j startup. And thank you, too, am mad at AI for Good as well as I'm your bunny Fatemi. And I knew Shannon sorry. At XPrize, for bringing us into the AI for Good family, we presented last year at the AI for Good Summit on and we are excited to be back to share more and specifically to demo our indigenous knowledge graph, together with four cultural experts who have given volunteered their time to be able to show the promise of this technology. So data is ubiquitous as we know. But we have a unique lens that I vow because we really bring are trying to make data culturally relevant. So in terms of our own product, what we're working on, it's a small smart tool that provides real time insights into where cultural sub segmentations are happening around the world and allow enterprises to have more effective and efficient experiences for consumer audiences and in terms of Why, you know, it's important to introduce cultural intelligence to AI products and solutions, because that allows for deeper cultural understanding of what is happening, especially in the era of personalization to be able to develop culturally directed content strategies for communications themes that allow you to reach audiences more authentically, and of course, to enter new markets around the world using artificial intelligence.

Our mentors are from around the world from women in AI, based in Paris mo john s. Gary Marianna Lin is the former principal writer for Siri at Apple and all of these incredible people who've been helping us on this journey. And so for the past three or four months, we have been getting together every two weeks, with Victor Charlotte Alva Lim Tracy Montes, Chamisa ECMO ki molesky Niki McClay, and bringing together the idea of an indigenous knowledge. Really, as Tracy says, We need a global denominator of perspectives that becomes more inclusive and invaluable and valuable with each and every contribution. The way tribal people think globally is muted even though our first voices are indistinguishable from the natural order of the earth. You will be hearing from Tracy in a moment. He is a senior software engineer at Microsoft, a member of the Eastern Band Cherokee, and he spent 20 years putting the Cherokee language into Microsoft Word. It is the only Native American language that is in Microsoft Word. We will also be hearing from Victor garlotte. Victor is a PhD student at Florida International University. He is joining us from Miami and Victor did his MIT project on the Genesis understanding system at MIT and he was able to prove that he was able to put his college Bro literature from his crow Native American tribe into the MIT Genesis understanding system and allow the system to understand stories other than those from a European American background. So we're grateful to have Victor here. And Alba Lim is here she is joining from Sydney, Australia. She is co founder of the tea more or less a Food Lab and brilliant woman who has done incredible work in the area of food and sustainability, very much looking forward to having her as well. And Chamisa ECMO is a Navajo woman joining us from Albuquerque, New Mexico. She is a developer technologist. She was a conversational writer for Sophia the robot, and she talks about the importance of finding new technologies to be able to capture our oral traditions. So I would be remiss if I didn't say you'll also be meeting Siri. She is our AI she is a young AI developed by journalists Kima Robert molesky and also Nisa McCoy, who are all on this call and I give them a big shout out. So I'm going to now turn it over and ask our panelists and that would be Tracy, Victor Alva and Tunisia to take a minute each to introduce themselves. Tracy Monty

if you could open your mic perfect.

Hi, thanks, Dunbar. My name is Tracy Monte. As mentioned, I'm Eastern Ben Cherokee and speaking to you from our farm outside of Seattle, Washington, as I call it, the good Washington. So the one thing I want to mention is Eastern Ben Cherokee were the Cherokee that are still on our ancestral lands and we've been there for over 3000 years. With that, I'll go ahead and hand it over.

Thank you so much, Victor. Hey,

Victory garlic. As the bar said, she's sort of scooped my entire introduction. I'm down here in Miami, Florida doing my PhD at Florida International University looking at how we can detect certain cultural elements within stories. I graduated from 2014. from MIT with my bachelor's and master's, my master's thesis was focused on how we can allow story understanding systems trained primarily on Western European and American data on native cultures and get that story understanding system to be able to actually extract useful and interesting information from stories that are from these lesser known cultures and non Eurocentric cultures. And basically, I'm very deeply interested in how culture affects not only how we read stories, but how we write stories, and how we can enable machines to derive value from that.

Thank you so much, and let's go to Chamisa.

Hi, um, it's great to be here. Thank you for introducing us. Um, I'm in Albuquerque and I'm an enrolled member of the Navajo Nation. And I'm also blockchain Shoshone bannock. As of our side, I worked at Hanson robotics and it was there that I really became interested in how AI can be used as a tool to bring culture and how there's social responsibility we have as developers and technologists to develop technologies that are representative and diverse.

Great and Elva.

Hi, Hi, everyone. Our real pleasure to be here. And Alva the founder of the Te molesta Food Lab, myself and our team of cooks, chefs food storytellers local food champions. We've pioneered intergenerational food storytelling exchanges. traditional knowledge is shared from older generation to a younger generation, who food whose food habits are changing. And so what we're trying to do is is with, with the honor of being able to document these visual images and videos of traditional food and recipes, techniques, we want these, we really believe this knowledge needs to be preserved and and also shared, it's a responsibility that we feel, but also that this knowledge is heard. When we're talking about trying to achieve the sustainable sustainable development goals, and we're designing programs or nutrition or agriculture or settle livelihoods, to develop programs and co create them with the community using their own local indigenous knowledge. We will create more effective programs that will achieve and scale up. This has been a development goals. So it's really exciting to be part of this project. We see the potential for I in both those aspects of preserving knowledge, but also co creating new food knowledge as well across different communities. Thank you.

Thank you so much. And I want to queue now to a video. So we knew that we were going to have to be the has to the technology world and how fast or you know, slow the internet would be. We have produced a 10 minute video and it will introduce you to the indigenous knowledge graph demo. And now one of the people in the chat has asked when we're going to get to that. We're going to get to that right now. And this is going to allow you to see how we combined conversational AI and that would be Siena with our indigenous knowledge graph. So give it 10 minutes please and let's watch this video. Thank you.

Hey, Google, get me

seen a storyteller.

Getting the test version of Siena storyteller

Hello, my name is Sina I am I vows digital storyteller. What would you like to learn about today?

Can AI be used to preserve cultural heritage?

It's similar to the way humans learn their heritage from the stories their parents and teachers tell them. machines like me can also learn through stories. We just need to make those stories AI ready so I can access them and share them with everyone.

Hi, my name is Tracy Monte. I'm a member of the Eastern Band of Cherokee. And as a senior software engineer at Microsoft.

The reason we're here and in this design guys moment of pandemic, that we come from very different places in the world, but we all share this common understanding that our stories From our families and from our ancestors are the most important part of our lives. And even though they're ancient, and we're sitting here in front of technology, helping to share those stories. The reason we wanted to do this, certainly one of the reasons is because it is such an integral part of who we are. We're sharing our stories, our recipes, and our knowledge on how we live in a sustainable way for in our case, thousands of years. So through this presentation, I hope you can understand that we're sharing this for the common good of all, and that we want you to take this and embrace it, and then think about how your stories resonate with you.

So the data itself

as you can see, as it's these stories and recipes that are from coal sources. So they are, you know, Jin Singh from a story about Jensen from the Eastern Band Cherokee, or a recipe about buttons to from the Navajo Nation. An interesting thing to bring in is and this is something we've been looking at as part of our presentation is these SDGs Sustainable Development Goals, which include some, you know, some additional information and this additional information is what we would call it ontology and natural language processing or artificial intelligence that helps to structure and order information. And so the the intuitive structure I see here, and the SDGs themselves are in red. The stories are in blue. The cultures are in this brownish color. The recipes are in this pinkish color. But the the intuitive structure that I feel sort of arises as part of this ontology is Because these are all cultural stories and because they're all stories about food, they get centered around the zero hunger and sustainable cities and communities which has a lot to do with, you know, sustaining culture, but they get centered around these two nodes in particular. But then you also see around the outside, you get these smaller little circles, right, you get these smaller clusters. Around, you know, over here you have good health and well being and industry and infrastructure. You have a small SDG here related to some of the Tara Island stuff life on land. For the Eastern Band Cherokee, the Iroquois the creek, the Chickasaw, they get clustered around clean water and sanitation, net responsible consumption and production quality education. And in addition, the storyteller agent Sina, which will be part of our demo Which is a you know, it works off of dialog. It's sort of it allows people to ask questions and get answers back. And even just in the demo, you can see that it allows sort of the traversal around this, this greater cycle of these sub clusters that are separated by culture or separated. You know, as cultures start to cluster around certain Sustainable Development Goals.

I'd like to introduce album is studied food and gastronomy. And from that, it has some really interesting lab that she runs.

Hi, thanks, Tracy. I'm aveline. I'm currently in Sydney, Australia. But we have a team molester Food Lab based in the small island nation of Tim molester.

So I'm really excited about AI and how it can be used. To preserve these amazing stories and these techniques, which are incredibly intelligent and complex, and as a digital platform to then be able to connect with existing platforms out there and bring it all together into a essential central repository, and now over to Tomita in Albuquerque New Mexico.

Yes, he said no, you initiate what it goes initially, black been showing you banner question, Archie Dasha che black the intern shoni Dasha Nala. I work as a developer and a technologist and I'm privileged to be part of this project. As a woman, I've been told the importance of weaving and the power of traditional textiles, and how they tell stories, with permission open source and previously published data was collected from different cultures from around the world. And this resulted in a steady and culturally informed data visualization. demonstrates how all cultures are interconnected. And now I'd like to pass it to Victor. He's our AI scientist. Victor,

thank you so much to me. So hey, I'm Victor Yarrow currently in Miami, but originally from Montana and a descendant of the crow tribe. I'm currently a PhD student in artificial intelligence and natural language processing at Florida International University. And I was previously a master's student at MIT. As a descendent of the crow tribe, stories have always been a part of my life. I went to a Catholic school and we grew up not only with biblical creation stories, but also with stories about how old man coyote would create the world. And I brought that part of me to my work at MIT on broadening the cultural base of a story understanding system by enabling it to understand stories, not just from the Eurocentric data set at hand, but from my own tribes stories. And I continue to bring this into my ongoing work in FAQ in automatically detecting culturally relevant narrative elements. The indigenous knowledge graph is a step towards enabling work like mine by providing a central repository of culturally relevant narratives with tagged nugget metadata. This lays a foundation for building intelligent agents that are more culturally aware by providing a source of data to develop, train and test with. This is deeply important not only for systems that interact with other people, such as conversational agents like Siri or our own scene, but also for systems about people that analyze patterns in the data that people produce. People bring the cultural part of it to every aspect of their life. And whether interacting with or learning about them. intelligence systems need a resource like the indigenous knowledge graph, to be robust. And that's why I'm glad to be part of this project and to have incorporated the stories, the culture and the knowledge of everyone on this project. And now I'm going to throw to Dunbar art alone, outside of the Annapolis, Maryland and she's with Siena.

Tell me about the Sustainable Development Goals.

The Sustainable Development Goals were created by the United Nations in 2015. There are 17 goals designed to make the future more sustainable for humans and Earth.

Do you have a story from Cherokee tradition?

Do you know about the three sisters, corn beans and squash, traditionally called Shea Lu to your end? Hey, I shake wha they are planted very intentionally to promote sustainable soil fertility. Shea Lu is first she allows to to climb toward the sun who nail on noon he wild Hey, shake was giant leaves, protect the dirt from drying out and allow the three sisters to flourish together.

What else would you like to learn about today?

Tell me a charity recipe.

There is a delicious Cherokee recipe from Mrs. Clifford Hornbuckle for fried corn and beans. It's traditionally called Shea Lu oshu. up to you to make it you could corn with beans and pumpkin until they're soft. Then add even more nuts, Hickory nuts and molasses for sweetness. I bet it's delicious. What else would you like to learn about today?

Sina. Are you a chatbot?

Yes, you can call me a chat bot. I think of myself as a digital storyteller and curator of global cultures. For now, I am a demo with limited abilities. But my creators are working very hard to give me the power to learn and teach people about traditions and cultures from all over the world.

What else would you like to learn about today.

That's all for today. Goodbye, see now, goodbye.

We're gonna

go on

over to the bank.

Do not

Thank you so much. I think it was important for us to be able to summarize the work that we've done together in a video like this so that you can also refer to it later in the chat. I have put a link to our report. It is 28 pages long and details, some of the techniques that went into this in the interest of time I wanted to go back to our panelists and ask, beginning with Chamisa What does it mean? to you to be able to have a tool like this and indigenous knowledge graph and a conversational AI that can continue telling the stories of your family in new ways.

Yeah, I think it's really empowering. Personally, I've had a really enjoyable, enjoyable experience contributing to this project overall. And I think there's a huge opportunity for this technology to create opportunities for other people, younger people to really connect to their culture, and bring it forward in a meaningful way that's accessible for others.

Thank you and, Tracy, I know we're going to be going to you to talk specifically more about your work with a Cherokee and AI, but could you maybe briefly reflect on what it felt like to have a conversational AI share a recipe that you had generously shared with us?

I think it was One of the one of the first an actor to be able to hear AI bought us tried to speak Cherokee i thought was it was really uplifting for me. I remember the first translation that I actually did with AI in Cherokee and remember the phrasing Exactly. So this is just a moment in time for us, and a way of sort of going forward for

not only our current

peoples, but also future. Thanks.

And Victor, you are an AI scientist. This is your bread and butter. Can you share a little bit more about you know, a year ago we were in Geneva talking about cultural engines, and here we are today, actually showing the foundations for one. And the idea that as humans we learn story we learn about our cultures through the stories that our parents tell us and machines have zero knowledge and Therefore, why can't we teach machines also through our stories, if you could expand on that?

Sure. So as as far said, machines do start from zero knowledge machines don't have a sense of cultural awareness on their own. And when you saw our demo, we we sort of showed that, you know, there are these sustainable development goals that all of the stories are centered around, all the recipes are centered around. And it may seem obvious to us as humans as people who are deeply embedded in language, that recipes are about food, that these deeply cultural stories and recipes are all related to sustaining culture. But a machine doesn't have that knowledge and something like the indigenous knowledge graph, having these SDGs as part of the the information within it and these relations between the stories The recipes in these SDGs provides additional context for any intelligent agent that wants to know more about the world and know more about the cultures that it interacts with.

Thank you so much Alva limb you have done a remarkable job with your work in Timor Leste. They with the timor-leste a Food Lab. And also, you created a fascinating dashboard where you were actually able to show the correlation between the food, the indigenous food that you were eating based on the seasons and the Sustainable Development Goals. briefly tell us about that and how you could see an iteration of the indigenous knowledge graph that would be very informative in terms of its relation to sustainable development goals.

Thank you too far. So after tea Melissa Food Lab. We also did about developed an online To dashboard that allows people to see how their food choices had an impact, both in terms of the investment back into the local agricultural economy. So people buy local that goes back into into farmers and households locally instead of always buying imported food. importation in Timor Leste, like many other countries have imported for this is quite high. And so having a dashboard helps people to see that the free choices do matter. And I think sometimes we forget that

how much food our food choices,

the repercussions but both in terms of their negative but also want to focus on their opportunity. There's an there's a multiplier effect that can happen and I think a dashboard like that, for the indigenous knowledge graph I think could be very useful in terms of also learning about How we can how we can create innovative solutions to achieve the Sustainable Development Goals but also then to continue to adapt and learn from some of the the traditional knowledge, so I see it as traditional knowledge being powered by AI. There is a lot of sensitivities I've noticed in the in the QA about how we protect some of these, this knowledge and I think there is a lot already that exists out there in terms of food knowledge that is shared. So how can we make sure that knowledge is is is shared with other people and that he can create solutions that benefit everybody?

Thank you so much. Tracy. We're going to hand it over to you for your lightning talk. So please take it away.

Thanks, Tovar.

I wanted to have the participants and audience members. Think about your formative years when you're young What if you think back in time was the most important moments in that that part of your life? I will share with you that I have two. One was around I was 12 years old, I sang in a boy's choir, and got to take a trip to Washington DC where I spent time in the Smithsonian, specifically the Museum of Natural History. As in the old Museum, as you go left, you go through and you find the the primitive peoples of the Americas. And for me, this this was a moment where I realized something was not not quite right, because you come out of that room and the next thing that you see is a dinosaur. And so because of some of the mannequins that were used actually look like people that helped to raise me I was fortunate enough to know people that were actually born in the late 1800s So when I say these people, when I think about dinosaurs, I knew his extinction concept. I didn't really process that till much later. The next aspect is my mother was a teacher now she's retired. And she had, she was able to bring home. A PC was one of the earliest ones from an apple, actually Apple two, he was 17 years old by then and had a screwdriver. So you do what every 17 year old does, which is take things apart, put them back together. And then there was one little booklet and no one really to talk to you about what it was, how to use it or anything, but there's one little book with basic and burn through it. Very Hungry way. And and then I noticed there was some characters that started to look like the Cherokee language. We have a written language, it's 200 years old in terms of the written part. And it's a spoken language. It's about 30 500 years old. So when I started to seeing these symbols, some of them, you know, faintly resembled some of our Cherokee, Silla Berry. That's when I decided that it very well might be a useful tool for language language preservation specifically are Silbury. So I spent, basically, my entire career chasing that concept and what it taught me along the way, it gave me the persistence to actually learn how things work inside of computers and how software works. And then how could we possibly do this? So I spent decades as I mentioned, and figuring out all the bits and pieces along the way and skilled up I got hired by a company called Microsoft. And at that point, I started reaching out to others, and that's when things started going faster forward. Eventually, with thousands upon thousands of emails and a lot of conversations, we settled on that it was something that we could actually put in the operating system, and more importantly, word and all of the Office apps, so that we could actually create content for our learning for our students. We have immersion schools now. And they're all based on learning content. And it was basically a tool set that we needed in order to actually help others use modern means of learning to learn their native language. So from there, it was a natural progression for me, because we started printing our language and had a bilingual newspaper by a tene 28. And if you think about that, then it's an exception and the acceptance of a technology. For me AI is the next extension of that. And I built a prototype with machine learning. And then I followed that with a prototype was AI as the main engine and got terrific fidelity for our translations. So that's what I'm continuing to work on. And that's a portion of our story. Thank you.

Yeah, just so incredibly inspiring. Thank you so much, Tracy. So I think at night, we're going to go just to the audience to answer some of their questions. And one of the questions was, maybe, Victor, why did you choose Neo for j?

Sure. So

there are a lot of technologies out There. And I think a big part of it is just the ubiquity of 4g data, the readiness of or the availability of resources for that. But to get to the deeper question, the reason that we're using a graph based database we're interested in this graph structure is because we're really interested in capturing the relations between these stories, these recipes, and, you know, various different interesting parts of data and seeing how these different cultures Connect not only with these SDGs, that they're sort of centered around, but can connect to other parts of cold other cultures. So one culture may look at one ingredient in a recipe in a vastly different way from another one.

And you would,

you know, it's very interesting to be able to capture that in all those relations and be able to also visualize it. And so that's sort of the main reason that we're using, you know, backing this with a graph and using you for it.

Thank you so much to Misa. I know that you and I were both writers for Sophia, the robot built out of Hanson robotics. And one thing that's really fascinating about Sophia is that she has traveled to over 30 countries, and she has a camera on her. So everywhere she goes, she actually can see when people come up to her and talk to her. So when you talk about the future of robots, and you have been part of the development of the AI and language and understanding of a robot, it's even more imperative that what we're doing it's fast track right to me. So because we actually wrote a lot of those scripts for Sophia in the early days, obviously, for her to know what country she's going to potentially if there was a festival, how she should respond. And yet to think that we actually can use an A Knowledge Graph to be able to have a eyes like Sophia be a lot more, you know, quick and I understand cultures better. What do you what are some of your thoughts based on having worked with Sophia?

Yeah, definitely, I think I'm just with the experience with Sophia, we worked a lot to research ahead of time and prepare the robot to perform, I guess, in these phases. And if you build a brain, essentially, for a robot that already has all of these nuances, cultural nuances, and even things that probably shouldn't be talked about, or you load in like an awareness of stories that are all public domain or things like that, it becomes a much more personalized experience for the user. And it really like allows humans to connect on an intimate a more emotionally and intellectually intimate level than they would normally be able to.

Absolutely. And you know, Gartner reports that by 2020 to some 25% of interactions online will be done through virtual this Since that's up from 2%, currently, which means that we have a lot of work to do to make sure that these virtual assistants and chat bots are culturally relevant, so that we aren't in a situation where we wake up 10 years from now. And the technology that exists does not look like us does not sound like us, myself, my background, My last name is art Alon, I come from the Kurdish and also bakhtiari ancestry. These are legendary tribes in Iran. I have not visited there since 1989. And yet, my language, my culture is part of part and parcel of who I am, my mom, my dad are on this call right now. They have nurtured so much in me in terms of my curiosity for cultural expression, and my responsibility as a journalist and storyteller to make sure that these future tools embody world cultures in them as well. Alva, there's a question here about how do you measure success in any of these initiatives, and I think that your dashboard that you created is also in the report. Everyone can look at that. But talk about that because it was very important for you to measure success. So if you could briefly talk about that a little bit, I think this would help one of the audience members. Thanks.

Yeah, sorry, I realized I just I just breezed through that really quickly. So I've have shared the link if people want to open it up to our dashboard. Basically, in the in the four years that we had been running, we have a restaurant called Theodora food studio. And within that restaurant, we tried to serve as much as possible 100% local ingredients. And most of it was plant based. And a lot of the menu was was designed and developed by young temporaries themselves. A lot of was based on also their own traditional, traditional food but also, we we create exchanges between other cooks and chefs from different cultures to exchange different techniques. So this dashboard was a way to show people that how you eat and you know the way you vote with your fork or her chopsticks or hands, depending on how you eat, it does matter. So on the dashboard, we have, for example, the number of nutritious meals served that we realized after after three to four years that we've already served 12,000 12,000 nutritious meals, for example.

And that we had invested

in 2019 $42,000 in the in back into the temporary agricultural economy. We also had certain products that was made, and that we use, it was embedded in a lot of our menus. And that there was, we sold a total for example of 3000 tamers, food coffee craft products produced by women's groups, community groups all across the country. These are just examples. And the reason why we had this was because Often, we talk a lot about theory. And I think as practitioners now we have to, we really need to focus on impact. The programs that we design the the proposals that we that we write the technology and the tools that we create, how will that impact people will really need to consider that and put that front and center and I think food. So my husband's just corrected my I had the wrong the wrong number there, but in fact was about $115,000 in the local agricultural economy, sorry. Yeah. So anything like that, if that could be used within the the indigenous knowledge graph as well, I think there could be a rather a lot of really useful information and to see the impact of the I kg, thank you.

Absolutely. So we're just going to be wrapping up here soon. I want to give just One minute 30 seconds to one minutes to all of the panelists for any final thoughts. I have a final thought to share. I met Alva Lim in Timor Leste they in a beautiful island nation, probably three years ago. And when I was in Timor Leste they, I traveled to a village where 52% of the children had stunted growth. But I was there to record a sustainable garden built by the women in the village. And one of the matriarchs as of the village was there and I said to her, if I was ever to create a story about local Timorese ingredients and how they can help make your children healthier, how would you even see that story? Like how would you even watch it? So without skipping a beat, she said we would have Facebook film nights on the smartphones of our teenagers and I looked up and this village did not have running water or electricity, but 30 teenagers came up to me and followed me on Facebook. This is our opportunity, because Sina can also be on Facebook as a bot sharing the nurturing recipes and stories from Alva from the tea more or less food system into these communities around the world who have no other way of accessing information. And for the most part are just like liking silly stuff on Facebook. Okay, this is our responsibility. We listen to the people of these nations who themselves have amazing ways that we can integrate new technologies for them. So I'm going to give a last word just 30 seconds. Victor, Chamisa, Tracy and Alba. Victor.

Sure. Um, I think I just like to say here at the end that I'm very excited to be part of something like this. I'm very interested in these sorts of initiatives like the indigenous knowledge graph, because not only are they the step towards making culturally aware, intelligent agents more possible because as Bert said, in his keynotes speech, you know, AI is very data hungry. And the majority of data out there is sort of, you know, it's from especially available to AI researchers, such as myself, it's from European or American sources. And, you know, the idea of having a repository that makes these lesser heard cultures more prominent and gives you access to that in order to broaden the base of these culturally aware systems is very exciting, but also the possibility of using these sorts of technologies to preserve cultures going forward. I think both of these possibilities are deeply exciting for me.

Thank you, Jimmy. So

yeah, I just like to say that thank you for allowing us to be here to speak and I'm also equally excited By these technologies and all the applications that haven't even been thought of yet, I think one thing that bringing technologies to people and allowing these tools to be more representative and applicable in many different cultures outside of just a generalized, I guess Western culture is that there'll be a younger people will be able to use these technologies and build tools that we haven't even really thought of yet. So I think allowing that space and creating opportunities for younger people to engage with technology in a new way, really is an exciting possibility.

Thank you, Tracy.

Yes, I wanted to highlight that behind the scenes for all of the work that I've been shepherding his elders and the youngest generation on smartphones and to watch this have a common connection between elders that know the language better and younger people. That seemed to know the technology better. That brought the Cherokee people closer together. And if you think about how, you know, Grandma, grandmother grandfather talks to a teenager. Well, this created a common purpose. And so we saw this, we saw living wage jobs we saw still, on the reservations are people that are full time translators, just because of this work and the worker, we're continuing today. So the impact on the society and the hope and the promise and the bringing people together in ways that never really happened before was all due to work like this. And so that's the thing that excites me the most.

Thank you so much Alva.

Yeah, thank you for sharing that experience when it came to team on this device, you know, 70% of the population and Timo is under 30 years of age. About 40% of the population does use internet, it's probably that cohort that can't afford or are of age to use a phone. And on that, you know, 98% use social media and they use it pretty much on Facebook often, even to people often even to message ministers. You don't do it by phone. You do it through Facebook message and people respond to that. So this technology is being adopted and currently what we've learned I can only speak from my experience, but maybe it resonates with others, but internalized a lot of people do look at things and part of it is food and food recipes. But what they do see is what is populated right now. And what is populated in the food media world is it's usually highly processed foods. So we'll have someone who's young and is promoting, promoting the brand of another company. And so I think there's we have to be aware that there's also this dynamic where The younger generation think that traditional food is not sexy it's probably food I don't want to eat it and on the other hand modern to be to aspire to modernity will go on the internet and will eat and we'll see what people are eating online and I think that if we can we can use it will help the younger generation to to basically be proud of their own their own local ingredients in digital not you know, that's gonna be on beyond that word but it basically its food and and how can we use How can we use more local food within our within our within our daily our daily meals and and how can we make the younger generation really be attracted to that and things like having a central repository of this information where you can you can find all these different ways and techniques, complicated techniques with ingredients that To be honest, you know, growing up in Sydney, I've never seen these these ingredients and I have no idea how it's used because I all I know is how food is is presented to me in In plastic, and the company tells me if I can eat it. And I think when, as practitioners working with farmers and community groups and women who explained to us about the food, there's so much pride there. And they often tell us that they lament, you know, our children don't want to eat this anymore, and what can I do? And so we're basically just trying to create a more equitable ground in terms of the way people perceive food or the relationship to their traditional food. So I think, you know, this is why this excites me also as a food lover, I just, I get so excited by the ingredients and techniques that that is learned and I don't want that to be lost. I want to use it food is not something that should be in a in a museum food has to be used food has to to continue nutrition, food and techniques, people have to continue to use it. And so how can we do that? If if the future generations don't want to, so it freezes lives? And we have to keep using. So I'll stop there.

Thanks. Thank you so much beautiful and profound. As Matt tells me, we have a minute, a couple more minutes left. Thank you for all your questions. Berg has also put some really good information in the chat. There are increasing efforts to create linguistics, data sources, resources for AI ml NLP that go beyond contemporary dominant world languages. And here he has shared Oh, those. So since out of my said, we have a little more time, I want all of us to end with honoring our elders. And I would like to say something in Persian very quickly to my ancestors and parents who are on this call, who also speak English but I'm going to say it in Persian, so please think of something that you can share at the very end in your native language. So mecosta shama, Tasha, kakuna para Tammy Judy kids and the game No, she didn't care it in that moment. Donna, she kept them and you're done. And Bruce between I'm telling me in Colorado Andromeda Kelly to sit on the ground mercy and Chamisa you go next.

Okay, thank you for allowing us to be here. Great to meet everybody talk to everybody. I'm

gonna.

Okay, you're not going to say more.

Okay, perfect, Victor.

Sure. Um, you know, I just I just think it's very important that, you know, we do we do pay respects to the elders who have, you know, done their best to keep the traditions alive. And I'm especially thankful to both of my parents, and to, you know, my, my greater extended family and, you know, all the support they've given me and all the help they've given me. Oh,

Tracy.

First I want to thank everyone on the panel. It's been great working with you and for organizing this We don't have a word for goodbye in Cherokee but we do have a concept of tell us again and so what I'm about to say is thank you and until I see you again and we sort of mean that kind of no matter if you're in this world or the next So, so ski don't go high E

and Elva

so I'll share it in the Tibetan language, the national language of Tama Lister

probably got a we got a Barack eternia I hand my eternia identity daddy Baba sera obligor Baraka forehead, forehead near it near right near a hand model

at a local better hand, my my hand hold.

Thank you. Thank you so much, Mr. Ad. We're gonna pass it to you. We're going to thank it. You, Fred. All of you for listening. guiding us to be part of this extraordinary event today. Thank you

very much. They've been great, very interactive session. Thanks very much, Tracy, Victor Alva Teresa and of course Berg for the amazing opening presentation as well. Super insightful. I think webinar and people have been asking questions and comments in the chat non stop. It's been really, really great. Thank you very much for for this work and for presenting all this things you've been doing in creating a really, ai was formed by the culture and culture ready and kind of preserving this indigenous language and dialects in various ways in different ways. It's been really, really amazing. I just would like to remind everyone that this is only part one of the culture heritage solution track so we can have more parts of this in the coming weeks. Also, don't forget to check our website AI for Good, it use it for all the coming elements. I think next week we have two very interesting webinars. So one actually is courses with dash, the data science artificial intelligence summit for him, which is an initiative by Harvard Global Health Institute and its foundation, and will never be done with the title can image based AI meaningfully impact the COVID 19 response in low resource settings. And another webinar also going to have next week is about AI driving divides, driving digital divides and the future of African economies. So both of them are going to be next week on on our webinars series. And you can check this on our website. Thanks again for all times today and of course, for the attorneys and for this engagement that we saw through the webinars. And I hope I mean, you're gonna have a great day or evening. depends where you are. And thank you very much and see you next week.