- 4 days ago
"As AI becomes increasingly multilingual, important questions remain about which voices, cultures, and forms of knowledge are being preserved—and which are being left behind. From enterprise communication to cultural heritage, what does it mean to build AI for a truly global world?
"
"
Category
🤖
TechTranscript
00:20Good morning, Viva Tech.
00:23Bonjour et bienvenue for the third day here on the main stage.
00:28I hope you're all feeling well and motivated for what's set to be another fantastic day here.
00:35We have a fabulous line-up of guests, presenters and, of course, moderators and panels for you.
00:42Extremely insightful.
00:43So I look to look forward to when we kick it off with a woman who is an absolute wonder.
00:49She's phenomenal.
00:50Her name is Aya Bedir.
00:52Now, just to give a bit of context as to what she'll be discussing with Spencer Rice,
00:57it will be lost in translation.
00:59So that's the title of this opening panel.
01:02Right now, as we all know, the most powerful technology in human history
01:07is being built by a handful of companies, although mostly for profit.
01:12Now, our next guest has spent her whole career thinking things differently
01:17and asking herself one question.
01:20How about building technology for everyone, accessible to everyone?
01:27She invented little bits, the little electronic building blocks that taught millions of kids to invent,
01:33used in 20,000 schools.
01:36She then shaped Mozilla's AI strategy, and today she leads Current AI,
01:41a $400 million-plus partnership of governments, foundations and industry,
01:49built to prove that public interest AI can exist at real scale.
01:54She's an inventor, technologist, a social activist, an all-round wonder woman.
02:00Please welcome to the stage, Aya Bedir, in conversation with Spencer Rice.
02:25Hello, everybody.
02:29It's a little early, but I'm not too worried because, first of all,
02:35this is going to be, as you probably all know, live-streamed and archived,
02:39so if you have friends that you want to have watch it later, please do.
02:43But the more important reason is that I think the people that really need to hear this
02:49as much aren't even in this room, let alone in this building.
02:54You know, for all the talk about large language models and data centers the size of Texas,
03:00one of the elephants in the room is that there's 8 billion people on this planet,
03:03and they all have thousands of languages, tens of thousands,
03:09or hundreds of thousands of cultures and subcultures.
03:11And so the fun question is, how do you build these monstrous things
03:16and not lose the bits that most people, most of the places, most of the time,
03:23is their lives?
03:24So that's, you know, that's, you know, it's not something that's going to get talked about a lot.
03:30Everybody's talking about programming and virtual engineers and all these other things,
03:35but the fact is that as AI spreads increasingly into everybody's lives,
03:42how do you keep those parts of it going?
03:45So we've got Aya here.
03:47She actually has, we're going to break a little bit of news here.
03:50She has some announcements, but I think we'll save those for the end.
03:54And I just wanted to, my question is really a simple one.
03:59How did this happen?
04:02I mean, how did current AI come about?
04:04Because it's not an obvious thing.
04:05It's not venture funded or anything like that.
04:07Where did the idea come from, and how did you end up in that job?
04:13Hi, everyone.
04:14Is this working?
04:15Yeah.
04:17Hi, everyone.
04:18It's indeed very early.
04:19I'm thankful for people that are here.
04:22It's very early for me.
04:23I just came in from New York.
04:25Just came straight in from Charles de Gaulle.
04:28So literally.
04:31Let's see if I'm coherent.
04:34Current AI was founded in 2015 at the AI Action Summit.
04:40So the AI Action Summit happened here in Paris.
04:42It was stewarded and led by France.
04:47And at the AI Action Summit, the French envoy to the president, Martin Tisney,
04:56came up with this idea that big tech is currently running and controlling AI.
05:04And previously, when we saw this pattern happen with social media,
05:10we as a collective public interest community didn't do enough to intercept this control of technology.
05:18And as a result, if we look around now, our lives have been fundamentally changed
05:22by how social media behaves, affects our lives, and how this technology has really altered society.
05:29And so the insight was we cannot let the same thing happen to AI that's happened with social media.
05:36But that in order to really intercept this trajectory for AI, we need to play in the big leagues.
05:43That there's a lot of really interesting work happening everywhere.
05:46But unless you really play in the big leagues, which means bringing together governments, philanthropy,
05:52and the private sector together to form what is public interest AI.
05:57And the idea was born at the AI Action Summit to form this brand new partnership
06:05between government, philanthropy, and the private sector towards building AI for the public, a public option for AI.
06:14So you're actually building, you've got, you're building technology.
06:17It's not just a talking shop.
06:19Yeah, it's absolutely not a talking shop.
06:21There's a lot of really good talking shops, a lot of people in civil society and researchers
06:26who have created amazing policy, amazing research, amazing content and awareness
06:32driving for what is AI in the public interest.
06:35In our case, we're trying now to take that to action.
06:38So we have a three-part strategy.
06:41It's called Fund, Build, Bridge.
06:44Fund is a grant-making program where we will fund groups, developers, nonprofits,
06:51organizations around the world that are creating AI in the public interest.
06:56That means AI that prioritizes people over profit and societal benefit over profit.
07:02Build is an actual building team where we are going to build a public option for AI.
07:10That means a version of a cloud, of a chat GPT, of an AI stack or AI product
07:18that is not controlled by any one company or country.
07:21And then the third one is called Bridge, which is about investing in this infrastructure of the movement
07:28so that we're not individual initiatives happening around the world,
07:32but we really coalesce and become a force to be reckoned with
07:35so that we are really a viable alternative.
07:39Now, when you say build, does that mean you're actually building data centers
07:42or are you somehow piggybacking on what's being done by the big commercial?
07:48So we are not piggybacking on what is commercial,
07:50but the goal is gradually to build up the entire stack
07:54from the hardware layer all the way up to the application layer.
07:57It will take time and it will take money, which is why current AI is well-funded.
08:02We have $400 million in pledges, which include $100 million from the French government,
08:11AI Collaborative, Prominent Philanthropy, MacArthur Foundation, McGovern Foundation,
08:16and also some leading mission-aligned private companies as well.
08:24I mean, can you also do, I'll use the word missionary,
08:29is there a way to kind of talk the big boys or big boys and girls into building their things,
08:37you know, with keeping these things in mind?
08:39Is that part of it?
08:44Maybe.
08:44I don't want to discount the important work that many people are doing towards,
08:48you know, trying to either regulate or trying to hold these, you know, big companies accountable.
08:53It's important work.
08:54It has to happen.
08:55I also think that you have to create an alternative to drive users to.
08:59When you tell users, you know, AI, proprietary AI is having harmful effects on education,
09:07on mental health, on, you know, social interactions, on creating isolation,
09:11if you're not giving people an alternative to go to, what solution are we offering?
09:17So that's, we're focused on the solution aspect of it.
09:21You studied, you were at MIT for quite a while, and you studied both sociology and engineering,
09:28as I remember, and one of the things I noticed is that you had a fairly,
09:32a sort of first principle of what you were doing with respect to looking at AI systems,
09:39and that is how does it work and who does it work for?
09:42Could you just sort of unpack what that means and how you see the current landscape?
09:48I mean, I think we all here have this, at this event, how AI works is reasonably well known.
09:54Who is it working for and how do you get it to work?
09:57I mean, obviously, you're trying to make something that works for more people.
10:00Do you think they're going in a harmful direction, or are they just forgetting about the other 8 billion?
10:10So, fundamentally, I believe that technology, and I've always believed that,
10:16technology is not a field or an industry.
10:19It is a layer that underpins all aspects of our lives.
10:24The way we communicate, the way we move, the way we interact with our own information,
10:30whether our bodies or our houses or our learning material, technology is underpinning all of it.
10:37And I've always had this belief that if a technology is so powerful that it's part of or controlling a
10:44lot of your daily life,
10:46then you need to be able to understand it, and you need to be able to have access to it.
10:50And access cannot just be contained to a few that are in leading positions and certainly not a few that
10:58are in Silicon Valley.
10:59And so, I believe that when the Internet of Things was coming online,
11:05it's around the time that I started my company, Little Bits, to break down what invention meant with physical things.
11:11I believe that when it came to robotics education and learning how to code,
11:17worked on breaking that down and making it democratic and accessible to people everywhere.
11:23And now I believe it very firmly in AI.
11:25If AI is the powerful technology we all believe it can be, and already seeing it is,
11:34we need to be able to understand it, and we need to be able to create it for ourselves.
11:40People, communities, you know, organizations, small countries, people that are typically underserved by technology,
11:50need to be able to create their own version of this tech that works for them,
11:55that really understands their problems, understands their needs, that respects their languages,
12:01that respects their customs, their cultures, their dialects, their food practices, their oral histories.
12:07All of that needs to be embedded in the way the technology works for people.
12:13It's not just a matter of creating a technology that is used everywhere.
12:18It's a technology that has to work for everyone.
12:21In a funny, I mean, the only, in the lives of most of us,
12:26sorry, the Internet revolution and the World Wide Web are the first thing that happened.
12:33And it really did, you know, you can go practically anywhere.
12:35I don't know what percentage of the people in the world are online.
12:37It's a lot. It's almost everybody.
12:40And one of the interesting things about that is that in the nature of the web,
12:44it's kind of bottoms up, you know, local culture.
12:47You can go on YouTube and see all kinds of stuff that literally you can't find anywhere else.
12:52It's just there.
12:53And it bubbles up from below, and it's keeping on doing that.
12:56And that has its own set of problems and algorithms and all those kind of things.
13:01It seems like the AI thing seems to be much more centralized.
13:07Top down, yeah.
13:08And top down.
13:09And that makes it a lot harder.
13:12I mean, the chances of things getting steamrolled to use the, seems like very high.
13:18Very high and happening as we speak.
13:22The example you gave is the perfect one.
13:25The World Wide Web is the most successful public interest technology of our time.
13:31It is a technology that did not need to be open and free.
13:38Tim Berners-Lee, when created the World Wide Web at CERN in Switzerland,
13:42decided to give it to the public, decided to put it in the public domain.
13:49And so the World Wide Web is free, open, available to anyone to use, and to build on top of.
13:56And through that, you had thriving businesses, creativity, cultures that represented themselves
14:03and created spaces for themselves.
14:05And it's a beautiful thing.
14:07AI, like you said, is top down.
14:09It's being created in a handful of companies in Silicon Valley
14:13and then being, you know, bestowed on the world in the way they see fit.
14:18And it's not only creating a lot of harms, it's homogenizing a lot of parts of life
14:25that are really important.
14:26Art, music, language, writing, thinking, and that's really problematic.
14:35If you ask those guys, you know, they'll all say, oh yeah, we consult, you know, we have
14:41policy teams, we have, we're very, it just seems like, given the amounts of money that
14:48are getting involved, that less, you know, the rest of the world is going to, how do you
14:56stay in that game?
14:57I mean, the amounts of money being thrown around are vast.
15:00I mean, you know, if you had the world's best fundraising operation you could possibly
15:06dream of, you wouldn't raise as much as gets announced every day by the big AI company.
15:13Every 100 million here, 100 million there, nothing.
15:15How do you stay in that game?
15:18We have something that I think is more important than money, which is people power.
15:23We have people, communities around the world that are saying, this dynamic doesn't work
15:30for us.
15:30Whether it's businesses that are worried about building their business, the technology of
15:37their businesses on top of big tech, and then realizing we don't have control over this.
15:41You know, any day, some of these big companies could change the terms, could turn off access.
15:47This just happened with Anthropic and Mythos.
15:50You can turn off access and then cripple a business.
15:54And so businesses are worried.
15:56Governments are worried.
15:57So now with a lot of the geopolitical dynamics that are happening in the U.S. and Europe and
16:01wars happening, there's a lot of worry, especially in Europe, but really around the world in
16:06governments about these governments not having sovereignty over their stack.
16:12And so they're worried.
16:13Then you have individuals and you have people, artists, communities, creator communities.
16:18You have, you know, grassroots civil society all over the world that also are worried.
16:24So there are more of us than there are of them.
16:27And our approach is to really create a platform through which we can create this global collaboration.
16:34And then our initiatives and our work can go farther.
16:39So to that point, you've got, why don't you go ahead, you've got a bunch of announcements
16:44of actual things that you're doing.
16:47Why don't you give us a little taste of what those are?
16:49So as I said, our strategy at Current AI is, we call it Fund Build Bridge.
16:55Fund is our grant-making program.
16:58And today is our official announcement of launching our grant-making program.
17:03Our grant-making program is called Culture Preservation.
17:07There's many people interested in expanding language diversity in AI, but we believe that
17:14language is not enough.
17:16We have to expand cultural diversity in AI and also cultural preservation in AI.
17:20So today we launched the Culture Preservation grantee program, the pilot cohort, and we're
17:28announcing the pilot cohort in partnership with the French government.
17:33So the French government is investing six and a half million dollars in this grant-making
17:39program to support AI that preserves culture all over the world.
17:44And this first cohort that we are launching today is for four grantees that work in Africa,
17:51in Brazil, and in the Middle East, focused on cultural preservation.
17:55And I can tell you more about each of them.
17:57Yeah, sure.
17:57Go ahead.
17:58I mean, so these are existing organizations that applied for funding.
18:03Yeah.
18:03And presumably through the current AI website.
18:07Is that how all this happens?
18:08So some of these, yes.
18:10So in the future, the grant-making program will be on the website and everybody will apply.
18:14The first cohort, just because the organization was being stood up, happened through individual
18:18conversations.
18:21But these four organizations exist, but some of them are really small.
18:25What are they?
18:27One is called AIRA, African Internet...
18:32Why am I forgetting?
18:34Oh, yeah.
18:35African Internet Rights Alliance.
18:36Sorry.
18:37And where is that based in...
18:39Based in Kenya.
18:40Okay.
18:40This organization focuses on developing accountability tools for AI.
18:47So a lot of AI that's being implemented in Africa is done with very little oversight,
18:53very little accountability, and a lot of bias.
18:55And so what they're doing is developing these methodologies and tools so that there can be
19:00a more rigorous and accountable systems.
19:03And they're sharing that work with communities around Africa.
19:07Then we have the Institute of World Making, which is based in Beirut, Lebanon.
19:13That's an organization that's focused on cultural preservation and creation in AI for the Arab world.
19:20So they work in Jordan, in Lebanon, in Palestine, in Egypt.
19:24And the focus is on how to preserve and how to respect Arab culture and language.
19:32There's over 30 different dialects of Arabic, for example.
19:36Most AI systems only focus on modern standard Arabic, which is one, and kind of erases the others.
19:41So they're developing data sets that are much more enriched, and they're developing tools and models and agents to support
19:52that.
19:53Then the third one is called Portal Sim Porteias.
19:57And they're an organization based in Brazil, and they're focused on how to create systems so that the indigenous Amazonian
20:07communities are able to create AI for themselves
20:11that really understands the nuance of their culture and their languages and their dialects.
20:18So if they want to, for example, interview an elder about ecological knowledge that's centuries old,
20:27how do you do it with AI in a way that really kind of like understands that material, respects it,
20:33and also is consensual?
20:35So they're developing all sorts of tools and technologies to do that.
20:39And the last one is called Masakane, as well based in Africa, and handles over 50 different languages in Africa.
20:49They're the largest natural language processing community in Africa, and are doing a lot of work to support the startup
20:57community.
20:58They're doing a lot of work to support nonprofit organizations and researchers to also create AI for the continent that
21:07is inspired by real use cases and real industry and individual needs.
21:13I mean, it would seem to me like, you know, all this is all great stuff, and we hope it'll
21:19work well.
21:21Is there a way, does anybody have a vision for how you could somehow coexist with the big ones, the
21:30large language?
21:31In other words, somehow or another piggyback or leverage or somehow, you know, rather than trying to, I mean, building
21:38an alternative is always a nice thing to have,
21:40but sometimes is there a way to jujitsu and judo them?
21:46Is anybody talking about something like that?
21:49Some people are talking about it.
21:51I think there's a lot of very, very interesting work happening around the world.
21:58The issue we see is that people are not connected to each other, so it's not adding up to something
22:04bigger.
22:05You know, big tech and technology companies in general are very vertically integrated.
22:10They, you know, their teams are large, and their resources are pointed to the same roadmap, the same mission, the
22:16same vision,
22:17and they're harnessing the power of that vertical integration to make more impact.
22:22We're trying to do the same thing in the public ecosystem across the world,
22:27and we think that we're going to get a lot of gains from these interconnections.
22:31I was going to say, I mean, one of the interesting things about AI, I think, that isn't really widely
22:36understood is that AI is probably the branch of computer science that was the most high-minded,
22:44the most, the least commercial, partly because it was completely non-commercial for the first 50 years.
22:50And the people in it are not, you know, many of them are driven, they don't even care about money.
22:56They are, they're the best people in computer science.
22:58So is there a way to kind of motivate them and have people in these big companies who do stuff
23:04that isn't doing all these harms that you guys are trying to prevent?
23:08It's already happening.
23:10We are getting calls and inbound all the time for people who are working at these big AI companies who
23:17are saying, we want to help and we want to work with you.
23:20And we believe many of them, when they see kind of momentum, will want to jump ship.
23:26We're seeing, like I said, a lot of government interest, a lot of industry interest.
23:30And so the momentum is building, the snowball is forming.
23:34We believe that the gains are going to happen quickly.
23:39We're very motivated, very excited.
23:42We're very grateful to the French government and the President Macron for taking a big bet on this ahead of
23:49anybody else in this space and really making a commitment to public interest tech.
23:57And we're, we're, we're excited where it's going to go.
24:00Well, then I hope some of you are out there that are similarly inclined and I'm sure you'd be happy
24:04to talk to them.
24:05So thank you very much for your attention.
24:07Thank you, Aya, for coming.
Comments