IGF 2018 - Day 2 - Salle III - WS172 Accessibility Improved: building inclusive societies with AI

The following are the outputs of the real-time captioning taken during the Thirteenth Annual Meeting of the Internet Governance Forum (IGF) in Paris, France, from 12 to 14 November 2018. Although it is largely accurate, in some cases it may be incomplete or inaccurate due to inaudible passages or transcription errors. It is posted as an aid to understanding the proceedings at the event, but should not be treated as an authoritative record. 



>> CAROLYN NGUYEN: Good morning, can y'all hear me okay in the back? Yeah? Great, thank you so much, everybody, for coming and joining us. Welcome to IGF workshop 172, which is titled Accessibility improved, building inclusive societies with artificial intelligence. For how many of you is this not the first session in AI at IGF? Okay, great. So, the majority of you, this is the first AI session. Perfect.

      This is organized by the International Chamber of Commerce business action to support the AI society. The association for accessibility and equality of Kenya, and also the Latin America internet association in Uruguay.

      So, I am Carolyn Nguyen. I'll be moderating the session today. Welcome. I'm excited to be here today to discuss with all of you about how AI can be used to address accessibility and inclusiveness. And so, what do we mean by that? It's about how artificial intelligence can empower everyone, regardless of age, ability, or language. And/or professionally by being able to integrate them better to the Bork force. So why is this important? 80% of them live in a developing world. There's a growing demand by people with disabilities and impairment and the trend would continue to grow faster due to the aging demographics. So, for example, in the U.S. with the aging of the baby-boomers.

      So, let's start with the basic. I want all of us to get to the same level on what is AI. And I'll define it quickly as a set of technologies that can help to improve the ways in which machines interpret the world around them. So, these exact same capabilities can be used to help people to interact more naturally with the world around them by reducing barriers for them in both professional and social settings. By creating new opportunities and helping them to realize their full potential. What I want to do is start the day by taking a look at the two videos as examples of some of the capabilities and I'll introduce the panel. We'll go ahead and play the video. Capabilities developed by Amazon.


     >> My son has severe autism and not able to communicate. As a technologist, I've always looked for different ways to help him communicate. So, he can speak certain words and phrases and he can read and write, but he can't hold a conversation with you. The idea behind Pollexy is to help him and others with dementia and other special needs to give him that confidence to not have to have someone constantly in their space monitoring them, but sort of invisible assistant that lets them have that freedom and sense of confidence to live their day-to-day life. It's connected to a speaker in his room as well as lights and buttons.

     >> Good evening, it's time to get ready for bed.

     >> The lights let him know that there's a message that is speaking and the button is a way for us to confirm that he's actually listening to the message.

     >> Please push the button.

     >> Typically, with people with autism, they love routine and they work best on routine.

     >> Good morning. It's time to brush your teeth. Okay, now brush your teeth.

     >> It makes him calm on a regular basis. It's a great tool to give him a sense of consistency that's really important for people with autism. The fact we have this in Lex and Polly and Alexa, there's support to build this. I feel encouraged and empowered to build this for my son. It's very fulfilling because maybe it's the tool that will reach out to him and break through. 

     >> That's the next one. Great, we have another video. This is from Microsoft.

     >> Where I live, they believed deaf people can't do anything. So, it was very, very difficult. I came to RIT because of the communication access that's provided here. RIT has nine different colleges. NTID, the national technical institute of the deaf is one of the colleges. It's the main stream program for deaf and hard of hearing people. The largest interpreting services as well as the captioning. It's the largest but we can't keep up with the need in services.

     So, we decided to use Microsoft translator as an additional communication to help on the journey to scale. Microsoft translator uses AI to create another grid for the gap that's been there for so long.

     >> As deaf person, I want the exact same information that may hearing friends have. Presentation translator was easier than we thought it would be. You really just have to click it and the software automatically reads the content and everything that you have within the PowerPoint system. The cognitive services custom speech recognition is critical for capturing vocabulary words that wouldn't be conventional in everyday life. Students can pick any language they choose to receive the information if the professor has chosen English, which they speak, then I can choose whatever language I learn in best.

      You guys play any video games?

     >> Students can use the app to initiate a conversation with others. So now that I have my phone, I can see exactly what was said.

     >> There are barriers to communication everywhere, but I think it's time to look at the barriers as opportunities and they can be broken down. Microsoft translator has the ability to provide opportunities to reach out to everyone.

     >> Great, thank you. That was the technology that we're going to demo for you, so we can run the session live --

     >> Going to be interviewed over Skype and you want to look better than this. Well, here are four easy things you can apply -- 

     >> CAROLYN NGUYEN: We have a little bit of a network issue because all of that is running live off of the network. I couldn't do it. Another thing I would like to put it for everyone to consider is all of the 170-plus countries that have signed or ratified the U.N. convention on the rights of persons with disabilities are required to put policies and practices in place. So, I'm hoping that we can use the conversation here to have a dialogue and explore how some of these technologies can be used to implement and increase accessibility -- improve accessibility for everyone, including, for example, one of the things that we started looking at is overcome barriers and access to justice. And I'll come back to that as part of our dialogue.

      So, in terms of the session structure, we'll do it as -- there will be two segments. First is interventions from our speakers on both technology and policy. And the second session is very much about an interaction with speakers and everyone.

      And to help us run all of this, because we always need that, the remote moderator is Sharida and Lyla. I hope I got that right. And thank you so much. And I'll go ahead and introduce our panel now. To my left, all the way left, is Gonzalo Navarro, who is from the Latin American internet association, the executive director. Next to Gonzalo is Nobu Nishigata, a policy analyst and leading the work, part of the team that's leading the AI work at the OECD. And then to my right, I have Chris Wilson who's a senior manager at public policy at Amazon. And then Ms. Olga Cavalli who's the advisor of Ministry of Foreign Affairs in Argentina. Can we have the other speakers remote? Okay. Perfect. Perfect. We have remote speakers who may join us as the conversation goes along.

      Can you share with us the initiative on AI? Just a tremendous sliver of the tremendous work that you've been doing.

     >> NOBU NISHIGATA: So do you have the slides up there. Hello, good morning, everybody. I'm Nobu Nishigata and the IGF and I'm honored. But today -- Your slides -- could you -- we started to work in 2017 after the meeting this year hosted in the Japanese government at that year. And then we had the meeting in Takamatsu in Japan where the G-7 leaders got together and started to talk on artificial intelligence in the future.

      Then, we had a couple of events and particularly like we had a huge conference. We had occurring as speaker. But it's 2017, last year, October. And then we had the joint show here in IGF last year and in Geneva. Then, now we are going to do the current work is one that -- the biggest work. It's American report with -- with developed in whole -- a report on artificial intelligence and the policy issues and the development and the applications. And the other industries.

      And then the other thing is we are now working on the development on the principles, which we'll have a trust -- there you go. I'm talking about the ongoing number two. Yeah. Thank you. We had just studied an expert group which will develop the principles about showing intentions which will force our trust and adoption in the society. Then introduction for the next year, we're going to develop a symmetry and also have that. And also, deduction and trust and society. Next slide, please?

      And I'll just introduce a couple of findings today, which is -- sorry about this. But we're going have -- we're going to talk about our report tomorrow, actually, in another committee with OECD. So just an introduction, a couple of things. One of the biggest trends in the investment to the artificial intelligence is the investment to the startups and the small company and those kind of things. And it's very, very rapidly growing as you can see.

      And the other thing is some location like China and the U.S. are very active. We tried to define the application of artificial intelligence and we got in the topic in this session. I would like to mention about the techniques, the combination between the actual interventions and the argument reality or maybe the data so it's kind of an advancement of the technology with that. You can -- maybe you know about the Pokemon Go, you can see then. And then you can -- with the combination of the artificial intelligence, which it has the images or the voices or the sounds, it has access to the sound like Microsoft one of the leading companies is the development of these technologies for the handicap people. So that if you see the glasses with these techniques, even if you're blind, the glasses are going to tell you what is around you. Or if you see the menu, if you know, then the glasses is going to tell you what's on the menu So that you can choose and you can order the and those kinds of things.

     Just introduction, I'll work on the policy development on OECD. We developed a framework. It's a portion that's studying the use of -- the use maybe like its application, even there for the handicap people. And the innovations, we need innovation to proceed in the technology use. And jobs -- some people know it, AI is going to replace you. We're not sure how fast that's going to come. But if you -- we had one report just to introduce that like the long-distance truck drivers, very specific to these challenges and maybe in that area we have to have some in these countries to -- for the -- for the drivers to give them new skills or just adjustment to the new technologies. Otherwise, once we have the drivers’ trucks and then it's going to happen.

     Other things, the policies, the main finding is right now, like more than a 20 or 25 countries already made AI strategies for their nation or developing national strategy, how to cope with AI and the G-7 study that discusses it in 2016 and the data continues the discussion and then after the G-7 meeting after -- sorry, the summit meeting. And then they adopted the document for the artificial intelligence society. Then in the G-20, Japan is going to take the presence next year for the G-20 meeting and they are hoping that they are going to proceed with some discussion about that showing intelligence.

     And for the United Nations like of course IGF but UNESCO started a discussion with AI and ethics. On top of these, the known government organization amounting the IEEE or the AI. Run by the -- including Microsoft, Amazon, and Facebook and Google. Like the people society, future life adopts principles for the AI. So, we have a bunch of initiatives and the policy leading papers about artificial intelligence. So, it's just analyzing these papers and studying working on developing the policy framework, how to cope with -- how to cope with artificial intelligence and what the problems could be in the future.

      The last slide is just an introduction -- this is the work study next year. So, we want to be the -- we want to create the windows because there's so many departments just, for example, talking about the government and then in countries like many, many departments are walking on that to some extent so we want to create a window to start a discussion and exchanging the information and best practice around the world. And we are just going to keep taking the stocks and the development of the policies and the initiatives and the artificial intelligence. And this is the end of my presentation. Thank you very much. Just introduction, our home page, very simple. OECD and AI just introduced our work. Thank you very much.

     >> CAROLYN NGUYEN: Thank you so much. I think it's fantastic that actually in the work that the OECD is doing to develop AI guiding guidelines are -- guiding principles, I would say that the first principle is inclusive and well-being. It's really built into the words. It's really, really nice to see this built in. And one of the things that I would like to explore in the next section is so how can we build that into the various different AI national strategies? But... 

      So next up, I would like to invite Gonzalo to talk a little bit about what steps can business take to improve accessibility for people with disabilities and some examples? Thank you.

     >> GONZALO NAVARRO: Thank you very much for the invitation. We're excited about the opportunity. Because obviously we're talking about the relevant topic for the future of our human kind.

      The more straightforward answer to your question is we need to do a lot in the future and in the present. Fortunately, I think the private sector is responding to this in the way of creating new technologies, devices, and service was for people with some kind of disability. And this is a crucial element because the beginning of your presentation, will have 1 billion people living with some kind of disability. And 80% of that 1 billion people lives in undeveloped parts of the world -- in America, it's one of them. So, we are talking about the official element here.

      So, what we can do? Well, some companies have started special programs designed to bring accessibility to people with disabilities. Most importantly I think private sector or the business community at least Latin America or companies providing services in Latin America are aware of the necessities and the potential that they have to create better conditions for the people.

      From our side, what we're doing in Latin America is conducting research on how technology is going to have an impact in the force in the future. That report is going to be prepared by next year and we are starting a special session on how artificial intelligence is going to improve the conditions of accessibility in Latin America. It's a key element that is going to reshape probably most of the gaps we have in Latin America. We still have a gap which is accessibility to new technologies, but the technology is there. It's ready, and prepared to and it has been. We're hoping to share more about this in some minutes when he can talk with the audience and -- about this presentation. Thank you.

     >> CAROLYN NGUYEN: Thank you, Gonzalo. That's a rich area in determining the useful AI in the workforce to increase accessibility. I would love to come back during the Q&A session.

      So next we would like to turn to our next two speakers, really, to talk a little bit more about what -- how can somebody's challenges be addressed through the policy questions and what's enabling the policy environment to encourage development of these types of technologies for accessibility? 

      Olga, if I may, I would like to ask you to address and share your thoughts on what policy principles can be put in place to encourage development of these kinds of technologies and also to capture the benefits that they can bring to empower people with disabilities?

     >> OLGA CAVALLI: Thank you, thank you very much, thank you for the kind invitation. As you were mentioning at the beginning of the session, with 80% of disabled people living in the developing world, you can imagine how challenging this is for developing countries.

      The problem with using technology, I was in another workshop this morning about smart cities. I had the same idea in mind, that the priorities that our governments have, unfortunately, saying more urgent things. So sometimes the issue about smart CDs or the use of technology for disabilities is not that it's not important, but sometimes lag behind a long list of other immediate problems we may have. Argentina has 30% of the population below the poverty line. They're not disabled, they're people who cannot reach food and work. That's a big challenge.

      So, at a national level, we have some regulations, in Argentina, we have regulations, mainly focused in the possibility that Prime have access -- people with disabilities have access to, for example, health, education, mobility, and buildings and transportation and all that. In regards, we have the disability in mind when you design the tools, the technology. If you don't have that in mind, sometimes it's not an issue of perhaps not caring for disabled people. It's just not having in mind at the moment that you design the tool.

      So perhaps with the government and with not for profit organizations, we've been trying to promote that web designers on the west side have those rules in mind, so, perhaps, blind people or people with sight disabilities can have access to that information. I would like to share with you an interesting moment, at this moment or maybe later?

     >> CAROLYN NGUYEN: Go ahead, please.

     >> OLGA CAVALLI: This is -- I care for it very much. We have helped them with the not for profit activities in the sector. We have helped them a lot. There is a library for the blind. They -- they they make the audible books with the normal books. There's a group of blind or low-sight people. They built an NGO. And they have in their library 56,000 books in Spanish and they're adding other languages including Portuguese, English, and German.

      The beauty of the project is that they told us that only -- they don't have statistics, but they think only 2% of the blind people know how to use this technology to read. Imagine someone who cannot see access in 56,000 books. The life of these people could change dramatically.

      Also, printing books in braille, braille you say? It's extremely expensive. And the books are huge, difficult to carry from one place to the other one. So, this technology, this platform can be used by anyone having a computer and some tools installed in the computer. But what they told us is only 2%, they think only 2% of the blind or sight-disabled people, can know how to use. It's not difficult, but it's not trivial. You need some --off need some education to some time with them to understand how to use the tool.

      So, what they have been doing, and they have received many prizes for UNESCO. We have helped them in finding funds from internet society and other sources of funding, they create teaching places for the blind people, so they can go there with their computers and they get to know how to use the tool.

      So, perhaps in some hours, they know how to use and they know how to access all of these books. They have built this not only in Buenos Aires where they have the main headquarters but also in cities around Argentina and they started doing that around cities around Latin America. And that is not funded by the -- by the government. It's just their -- their ideas and going around.

      And I will finish in this comment, which I think is really relevant. They had trouble with electric property laws in the point that there's someone coming and saying this content is just not allowed. So, they started a work that ended with an b international tree t -- treaty that allows this content to be used for humanitarian reasons for disabled people. So, they have the right to use the contents for people that are blind, or they have problems. Not only blind. Imagine old people who cannot hold a book, or they're sighted -- they're not blind but they cannot see well, or people who cannot even move, maybe, it's easier to hear a book, an audible book, than read a book.

      So, imagine they started with this idea and it ended up building this international treaty that has a huge impact in the whole humanity having this kind of disabilities. Thank you very much.

     >> CAROLYN NGUYEN: Thank you so much for that really, really aspiring example. And also, for --

     >> OLGA CAVALLI: Just to note, the project is called -- LIBROS. I can send you a link. Anyone b can access it, by the way, not only my people.

     >> CAROLYN NGUYEN: That's great. You set up a very great example there. A couple of things that you put out there is really great for everyone to think about is diversity by design. Accessibility and diversity by design or inclusiveness by design is something you put out there. The other thing you made clear is that when we start to think about sort of accessibility and these kinds of policies, it's part of an oral holistic policy framework. Because what I'm hearing from you almost is an affirmation of the policy framework that Nobu put out that says you have to have accessibility in terms of being able to access the services, the capacity building, the equipment, etc., but also intellectual property. And also set up really nicely with audibles and also, with the value of partnerships.

      Can you talk about the kinds of applications people can have? And what steps are needed to ensure investment in these kinds of capabilities and applications as well.

     >> CHRIS WILSON: Thanks, carol. Good afternoon, everybody. It's a pleasure to be here to talk about Amazon's efforts on AI; specifically Amazon's efforts on AI and accessibility. I must admit, I'm a bit torn. I'm a -- obviously given a few minutes to talk about this initially. On one hand, I don't know what I can say that improves upon the videos you saw because the real world tangible experiences speak more loudly than I could say. Because AI accessibility is important to Amazon, I could spend the next two hours talking about what we do in our efforts. So, I will find a happy medium and speak a couple of minutes to you about our efforts.

      I think perhaps before we get into some of the policy thinking and investment thinking, I want to say a little bit about how Amazon approaches AI and accessibility specifically. I think it touches a little bit of what Olga spoke about in terms of accessibility by design.

      So, our mantra at Amazon is number one, focus on the customer. We work backwards from there. Whether we do, AI related or not, is to focus on what is good for the customer. Does this improve the customer's access to goods and services? Does it improve their lives? Does it make their lives easier? That's certainly true with regard to accessibility matters. Every day at Amazon is day one. We learn from mistakes. We see how we can improve upon that. You saw a taste of that with the PLEXI project and how to utilize Alexa to improve lives of those of in that case autism and other disabilities as well.

     All of the devices that we see as a -- as a true opportunity for the company and for customers to enjoy their lives and have a variety of goods and services that they might not be able to access. Facilitate access to services. Voice view. Voice view is a speech-to-text learning technology we use with our Kindle devices and our Fire tablets to allow the blind to access material on their Kindle that they wouldn't be able to access visually. That's what Olga was referencing. You have AI not as much accessing goods and services, but AI facilitating someone's life and helping in someone's day-to-day life.

     So, we see at Amazon that there's definitely a continuum. That equation if you will. Speak briefly. A few more echo devices we have. AI technology is for the hard of hearing. Alexa talks to Alexa which allows those who are hard of hearing or deaf to still use it, in this case, an echo show device and allow them to visually utilize Alexa where they couldn't necessarily hear and speak to her. So, these are all examples of -- small examples of how we incorporate all our customers and our thinking.

      The partnership on AI is one glowing example of where industry and other academics, etc., come together to develop for lack of better technology best practices and constructive thinking about AI. Those are the discussions. Proud -- how to engage in those efforts to try to think in part to hopefully the notion of how AI is and debunk myths about artificial intelligence. We see artificial intelligence as a technology or a set of technologies that should be embraced. Far more positive and good that comes from them than negative that you may hear about.

      So, I think the industry and other interested parties can come together to talk about AI, educate people about AI, educate governments about AI, I think we'll go a long way towards providing that enabling environment for the continue to innovate in this space.

      And I would say in terms of policy, etc., I know it was mentioned earlier, a lot of governments have launch initiatives on AI and looking at AI. And I think any discussion about policy for AI or any technology for that matter should sort of -- wants to focus on how can we enable AI, not how we can constrain AI or any other -- the more education that can be done, again, the more others can come together and work on best practices to inform the discussions I think will go a long way towards enabling further innovation down the road.

     >> CAROLYN NGUYEN: Thanks so much, Chris, for that. And also putting out there the importance of we all think about AI on really how AI can empower or augment, you know, human capabilities and create opportunities like what we're talking about here.

      For this next portion of the discussion, of the session, what I would like to do is just pose general questions to the speakers and actually anyone in the room as well.

      I would like to you know sort of have this part of the discussion as three-fold -- three parts. The first one is I would love to hear from -- from our experts here as well as anyone else in the room.

     >> One of the friends Veronica not in the room are doing would study the use of artificial intelligence for healthcare access, specifically diabetes monitoring. And for making -- and more detecting diabetic glaucoma. Like eye technologies. So, healthcare is one such area that human intelligence is making a lot of inroads in. And there are a lot of questions of bias that she is studying. So you should all speak to her if you do come across her at some point

     >> CAROLYN NGUYEN: None of our experts --

     >> I want to comment about long-standing challenges. Chris mentioned something important that regulations should see all of the technologies on artificial intelligence but on-line technologies for disabled people. I know some concern about privacy and the devices that Chris was explaining a minute ago. Some people are afraid this information is kept somewhere. So, people are starting to also think not only benefits. So, regulations should not be focused on those potential problems but focus on the benefits that could bring to the whole people, the whole society. The regulators, accessibility by design when they sign the regulations. They have all of the information so they can do when they sign the regulations and not constraint the good use of technology.

     >> CAROLYN NGUYEN: Thank you so much, Olga, for that.

     >> Yes -- it's true -- sorry. Regulators have to think and they have to regulate everything because it's a common standard. Never, never go at that that the innovation is going and the things that with the good opportunities that we see in the -- in the artificial intelligence can be extended to -- basically any technology applying in the region. So that's why it's so important what Chris was mentioning in terms of public-private initiatives. I mentioned Latin America in terms of accessibility. People that are not able to go to their work or to perform oh in that manner are going to be impacted in another way.

      So, what we need to do is to work towards that common understanding on the benefits and not only to put an eye on the possible consequences of the assistive technology and to bring out -- the platforms of any size to do what they do best, which is connect, offer, and demand.

     >> CAROLYN NGUYEN: Susana is remote. Can you hear us? Can you speak? Hello, Susanna? You're still on mute. So, we can't hear you if you are speaking. Okay, Susanna, if you can hear us, if you -- hello? If you can hear us, if you would like to make a comment, if you're able to type it to the chat window, we'll ask the remote moderator to read out your comment. We'll wait for Susanna to come back to us on that.

      So, I want to take up this point in terms of regulation, principles, etc. To encourage more the benefits of AI while acknowledging the risk. Nobu, I want to come back to you. So, how in the work that you're doing, in the AI initiative at the OECD and in thinking about crafting the principles, how do you think the points can be addressed and sort of clarity in terms of, again, how can the benefits in the potential of AI be maximized but while making sure that the risks are acknowledged and mitigated?

     >> Okay. AI is going to provide the tremendous benefit for our lives in the near future and the some of them already showing up. But some -- raises some issues. Of course, we do care about privacy. But before that, I -- let me be clear that OECD is not going to have any intention to regulate or nor -- no intention to recommend member countries to do some revision. We need more like innovations with the new technologies, just like you think going back to the where we had the internet for the first time, like the late 1990s, we saw the mini changes. We're not sure how far or how much impact AI is going to bring us. But we're looking at the new technology in a different way, though. But they're going to bring that huge impact on the improvements in our lives.

      The handicap people issue today is a good example like internet, of course, we had new implemented accessibility for the internet for the blind and the deaf people. But on the other hand, this artificial technology is -- AI technology is going to work for your ear or like a they work for your eyes. So, then, it's going to open up the new possibilities in which we did -- of course, we had expected but now the expectation comes true, now.

      So, in essence, we want to have more emphasis on the good side. On the other hand, the -- (indiscernible) we have to think like if you think the biotechnology, which is very -- strong cutting-edge technology as well, you have to think, we have to be very cautious sometimes to treat the new technology. Open up new possibilities to cure our diseases and the healthcare and those kinds of things. But, I mean, the technology -- we have to be careful sometimes. So, in that sense, we have cautions to the new technologies of the government as well. But my point is we have to be cautious to the new technology. You can tell by history we'll develop the technology the same way. But on the other hand, we want to put more emphasis on that side.

     >> CAROLYN NGUYEN: Let me get to the objective of the work, right? Which is a predictable and stable policy environment fastest trusting an adoption of AI globally is essential to AI that with inclusive a sustainable group and well-being. The focus is the recognition, the trust is an issue. That there needs to be -- that the principles need to work together to enable that trust. Chris' point, Gonzalo's point as well the objective of the principles is to realize sustainable growth and well-being. Sherida?

     >> There's a comment -- I just want to read out a remote comment. Just one second if I can get the thing to scroll. I'm autistic myself. ITU and the world health organization just set up a focal group on AI for health. We have participated remotely in and -- could I read from that one because this one is refusing to scroll. We have participated remotely in the health track of the ITU's AI for good global summit and the first meeting to focus group and requested to be informed. But any activities they may plan with autistic persons, persons with disabilities in general, and/or mental health. So far, we've not been added to the mailing list. What have can we do to ensure that persons with disability are included in the AI activities like the focus group. 

     >> CAROLYN NGUYEN. The organization that Chris mentioned before, the partnership on AI has a number of be civil society organizations as well as academics and industry from all over the world. Right now, there are more than 70 companies from pretty much different parts of the world. So, there are -- there are different ways for you to get involved. And let us know. Send us -- send us an e-mail, etc., and we're happy to incorporate you into various different discussions. Any MAK members? Ex-members?

     >> There's a push to have IDF for this particular topic on AI accessibility.

     >> That's for sure. You know, this is a continuous effort because since I came into this environment, I mean, like I don't want to mention how many years ago, many years ago, the situation is not improving multilateralism. But obviously the IGF was a tremendous contribution in that way because we're bringing different voices.

      When you mentioned that what we need are a set of Universal principles to be applied to the artificial intelligence, it's quite correct. That's the way we need to move.

      I will say that the mission of coming into control and the set of regulations not going to work especially if you don't have active intelligence or the new kind of technologies we're going to develop in the future. So common understanding and different voices bridges the dialogue and brings a better set of principles and thinking that we need about this issue.

     >> CAROL NGUYEN: I think also by participating in these kinds of sessions and being part of the IGF platform, I think that's another way to participate and develop an increased awareness of these kinds of issues as well.

      I want to come back to a point that Olga had made with respect to -- you know, when you're starting to look at a policy environment that would enable and encourage accessibility by design, you also brought up intellectual property. Can you talk about what -- how -- what are some of the other elements that would need to be there to create this sort of amazing environment, you know, I hear also capacity building. Gonzalo, I hear we need to focus on workforce issues, so Nobu, I hear from you access, etc. So maybe we can have a conversation around that?

     >> Thank you. I think for regulators, it's clear information, clear information about the benefits. There should be more promise that technology may bring, but that should be not even in the first line of the document. But I always have a feeling that there's a lot of information, very important things that the regulators and the government officials don't have on hand. At the moment they have to make their decision.

      They usually have so many things on their table to care. And that is, again, the real challenge in developing economies. The sense of urgency is big. And urgency that impact the whole society. So that is challenging. But captures the attention b with clear information and the clear benefit for some part of the society. Let them know that can change also in what happens in other cities and countries around. What happens in countries in Latin America is clearly impacting other countries very quickly because of language, we share the same language, Portuguese and Spanish are quite similar. So it's easy this investigation spreads all over the region quickly.

      I'm an engineer. We go with it, explain it, we're not so clear about benefits and clear concepts on how to target that easily and quickly. So that, I think, it's not only privacy, the intellectual property, but have the right information to the right people at the right moment, which is not easy.

     >> CAROLYN NGUYEN: Anyone else? Okay?

     >> There's a comment in the back.

     >> CAROLYN NGUYEN: Yes.

     >> Sorry, I came in late.

     >> CAROLYN NGUYEN: Hi, sorry. Can I ask you to come up to a mic so the remote people could also hear you?

     >> Hi, I'm doing a master's thesis on perspectives on AI. I just wanted to bring up, I'm not sure if this was discussed in the session, about how gender bias is coming so strongly when you design AI, especially when you don't think of it at the get-go from the planning stage, right? If you're bringing in someone towards the end as someone who will just take a look at whether the system has certain, you know, effects or not, it isn't going to be very effective.

      So, from the planning stage, you need to take a look at whether the algorithms that you're using are representing in your data all the populations that are going to be affected at the end of the day and really not think of technology as neutral. Because it does exacerbate the equations that already exist in society. So, if you're going to bring in an AI, that's going to make decisions, everyday life decisions for people, those decisions are going to exist in existing power equations, in power dynamics, and social and cultural dynamics that already exist.

      So, for example, a lot of the newer algorithms that are coming up such as deep learning, they are already efficient, if you define efficiency as how much accuracy you want. But if you define it as how much of the algorithm are you really able to understand and are you really able to say how the algorithm came to that decision, then it's not the best option to use, right?

      If these social implications of algorithms that really don't get spoken of much like in technical groups. So, I really think that's something that's important to discuss and keep in mind when we're discussing AI and when it's social implications.

     >> Thank you.

     >> CAROLYN NGUYEN: Thank you for that comment. I'm going to call on her a little bit, Layna? So, to the point that you're making. First of all, this conversation is very much about accessibility and using AI to be able to adjust marginalized groups, right?

      The point you're making, I want to turn it towards disability. I want to make sure I address it. You're making the point that technical people are not looking at bias. I would say there's a large research community that is addressing this question and we have, actually, a researcher working on that in the room. I don't know if she wants to make a comment or not. It's up to you. The comment is that the technical community is not aware of bias and not addressing them. And so -- so first of all, this is Layla from Microsoft research in Montreal. She's data scientist with our research organization. As you're making a comment, if I can also ask you to tilt it a little bit back towards the accessibility and inclusiveness conversation?

     >> Absolutely. Within the research community, there's a lot of concern about bias in the AI models and a lot of work on this making sure that we build data sets and we train our modems on data sets that are representative of everybody and that includes groups that need different technology for accessibility reasons. A new topic and an emerging topic for sure. There's a conference about this, fairness, accountability, transparency. And it's sold out every year. It's been happening for years. People in labs are worried about this problem and making AI accessible to everybody. So, there are technical solutions that are being proposed for those problems in particular. One is data set collection. Making sure that different groups are represented equally. Also training our model so they take into account equality considerations so that our metrics are not only performance, but also fairness of treatment of different groups represented in the data. So those are the approaches and the -- there's definitely a concern within the community and there's the -- there are definitely best practices that are emerging and that everybody -- I can say nerve the community is really applying those best practices and trying to enforce them so AI is accessible and usable by everybody.

     >> CAROLYN NGUYEN: Thank you so much. Yes, please?

     >> I'm from the Dutch government. It's really great to hear that this is happening, but I also have a question for Chris regarding Amazon. First of all, thank you for joining the panel. Great to hear that business s are concerned with these topics, rather than just NGOs and governments.

      In regards to building an inclusive society, I was wondering if Amazon tries to include the people who do not have the resources or the knowledge to use AI? And the underlying question here is how do we prevent the increasingly growing divide between those with resources and those without resources to use AI. Does Amazon see this as a responsibility that affects them as well. Or do you think governments and NGOs need to step up here?

     >> CHRIS WILSON: It's a collaborative effort. NGOs covers -- certainly the private sector has the responsibility to help. Bridge that divide if you will. Wide ranging initiatives to touch on that. Sort of hard to drill down now. AWS has efforts to do some capacity building in various parts of the world, the community to help try to bring -- to educate people and bring them up to speed for lack of a better word, on the benefits of technology. Everything we do is a customer. Not just a customer, certainly in the developed part of the world, all potential customers all over the earth.

      We are -- our public mantra is the earth's most customer cent Rick company in the world. That includes everybody. I'm happy to talk to you after -- point you in the right direction for more specifics. Suffice it to say that business at large is always trying to -- to -- at large trying to fill the gaps and help with the private sectors and NGOs and others because it's a team effort. Not one entity has a role. Thanks.

     >> CAROLYN NGUYEN: Thank you, Chris.

     >> Hello, I'm Cat Duffy with Interness. We wanted to talk about the program we're running in the open source development and digital safety tools. It's a bit removed from AI, but the structures I think are the -- they're potential sort of models that we've built that I think have been effective in this case and might be scaled as well. We have a site called useable dot tools which is available and on that site we have numerous open source personas that have been developed for representatives of disabled and marginalized populations all over the world. We'll have a blind LGBT rights activist. Right? We have a disabled activist from Mexico as one of the user personas.

      So, it's a wide blend. It's not a lack of development on inclusion but there are vast gaps in communities and we don't have a lot of mechanisms for bringing disabled communities, marginalized communities, high-risk communities together with developers. So, one of the ways we've approached this is to do annual meetings where we as a trusted interlocutor brings the groups together and we bring multiple tools together with populations that represent high risk for marginalized groups around the world for multiple days or a week to really work together and talk to each other and begin to understand each other. People coming in with fewer resources than the Microsofts and Amazon's. If that's an assistance to everyone, it's a product that we're going to continue to run and we're going to be happy to talk about. That's usable tools.

     >> Thanks. That sounds like a fantastic and offer to bring the stakeholders together to have meaningful conversations. Yes?

     >> We're talking about inclusion all the time. Which is great. What would be the possibility for businesses to consider with disabilities in their designs with AI, for example? One of the things is designing people with disabilities also benefits all of the people. So, it benefits the company and the public, but any other possible incentives for companies.

     >> The broken record, I'm getting back to Amazon's perspective; all customers are customers. We think of how does it affect the customer's life and his or her access to how to provide them. You mention that the notion of accessibility by design. We say it and maybe true for Microsoft. We think ant that through that lens. We're not perfect. We might miss the mark. Certain technologies not there at the time. We're trying to constantly think about ways to improve on that. The incentive is there because we want to serve all customers, simple as that. We don't want to exclude anyone.

     >> I'll add on to that as well from Microsoft's perspective. So, our mission statement is to empower every person and organization on the planet to achieve more, because we believe that everyone has value. So, accessibility is a pillar. Every product we need to make and ship needs to have accessibility capabilities. That's why you saw the development, for example, it's a power-point plug in. Actually rated -- the notion of performance, right, that is a part of our work. So, it's built into part of our work as well because everyone does have value.

     So, the ability to bring in to everyone to participate would believe enable us to build richer products, and as you said, it's not just for people who are disabled but it is a much better product for everyone. This year we launched AI for accessibility. So, the AI for good program is something like $115 million effort over five years. In terms of AI for accessibility, it's about encouraging. We encourage smaller organizations, anyone, who either have a need or an application to come in and apply for grants or to be able to use our resources to develop applications.

     >> I'm a student at U.C. Berkeley. I've got more of a comment than a question. Inclusiveness is how developing countries uses AI and how it affects their children -- their citizens. And one example of how to use AI grow food, what opportunities could be driven from India, how they use face recognition to spot 3,000 missing children by tracing them. And this is a great example of how it can be used. And as we also think of how normally systems, algorithms are trained by adult data, having the success of finding children is I think great. Thank you.

     >> CAROLYN NGUYEN: Thank you so much for that excellent point. We completely agree. Everyone on this panel completely agree with you. Yes?

     >> I'm a technologist. I don't work on this issue specifically, but to my community, people with invisible disabilities. And there is a gap in diagnosis with people with invisible disabilities. Diagnosis about ALS is very vocal about the fact that it took -- it takes on average 17 years with her particular disorder to get diagnosed because she gets dismissed particularly women -- particularly women of color don't get diagnosed with these issues. Because doctors discriminate against them. So if anybody is working on this, I would love to see research on the language that the doctors used to dismiss patients because we track when disabilities are diagnosed, but not necessarily when they're not diagnosed.

      And to that point, people with disabilities are discriminated against in employment, which is one of the reasons why privacy is such an important thing to consider. There are lots of benefits. But if you're a person with a disability, you poked and prodded enough. You have to give up privacy enough. So, uploading things to the cloud in order to enable accessibility, that -- that is a concern.

      Last point I'd like to ask the companies, giving grants is amazing, thank you. But how many people with disabilities do you hire to work on the projects? Are all these people who work on the projects on a permanent basis neural and biotypical? Are you hiring people with disabilities, not just to work on disability issues but to work as develop developers, coders, all those things?

     >> CHRIS WILSON: Excellent point. That's the first I've heard the term "invisible disability," I've been educated and that's important to me. And I think to the company. So, they contribute to make a difference. And I can say we welcome all I think I'm -- I'm not privy to specific hiring practices, but I'm happy to try to find that information for you.

     >> Thank you very much. In connection to your comments, I think you're right. There are some aspects that goes in terms of technology and we -- we have other issues that we need to look at in terms of privacy, for example. The question is how do you find the right balance in order to be -- and to do that at the same time. So, I think some standards are world-wide. And we are still working in the implementation of that standard, some countries. The present -- the artificial intelligence going to bring more spaces with people with disabilities. They're not going to -- not going to move, you know, to the work or the -- perhaps they'll have special conditions even if the -- (indiscernible) and in many countries, you have -- you have special in terms of borders or the number of people they had -- people with disability. And obviously, the technology is going to -- (indiscernible) so I don't have specific numbers, but I'm pretty sure that the initial findings of our research that was mentioned at the beginning is pointing in that direction.

     >> They do hire -- as he said, hire everyone who's qualified. And that's why we want to make sure that the developed technologies can enable people who are, you know, still participate and that's a major effort.

      And what I would like to do is go around -- are there comments? Sorry?

     >> There's a comment. Work close with organizations, public sector and the academy where we have the business -- I have tried -- okay. And also -- the sectors participate for all important perspectives to be covered. Many talk about the Universal design, but often it's no more than first do most of the design and then after the users are tested spend time and money on the project. You might actually say something after they've given the input. They believe in really starting with and -- the users. Workshopping, discussing, etc., the beginning of the project. We are actively recruiting people with disabilities it is for working with us. She can hear us.  We can't hear her.

     >> CAROLYN NGUYEN: Thank you. In the time we have left, I would like to ask every speaker to give a final remark. And what I would like for you to do is one take away and one call to action with regards to AI for accessibility. I'll start with you, Olga?

     >> OLGA CAVALLI: I was thinking about in the developing world, we consider consumers and developers. That's the fact that technology is mainly developed by companies which are -- most of them are based in the American economy. So, my comment for here with the a group of companies related is when they -- when they initially -- the community when a company comes to a developing -- they want to consume technology. Well, you have to inform the community. What do you want to be doing? And the opportunity to try and build a development center, maybe the government and try to build the knowledge and the local level. But tell that to the community. If you don't express it and you don't let the community know, people usually react negatively and is it -- it's that the companies want to come and -- (indiscernible) but at the same time, if companies could partner local to build local knowledge that -- and explain that to the community. That's a common user of technology.

     >> CAROLYN NGUYEN: So -- thank you for that.

     >> Build on that. A twofold. All technology. AI. How it can help people's lives, vitally important but even in the developed world, I think there's to educate, not just governments, but also the sense about -- the beef that's about -- my point -- the Amazon perspective, I think when we conceptualize that product or service, if we can't articulate why this is good for the customer, why this helps them or her, then it's not worth pursuing. And that's true with the accessibility issues and true on how to utilize AI and other extremes. When you think about these products, you want to use specific markets or use them globally, I think hopefully the industry and purchase things that way, take very seriously the impact technology has on people's lives. And we want a customer to feel so secure and trust -- trust our products and services.

      So, we have that and try to improve on that. And hopefully we can continue to do so. Thanks.

     >> CAROLYN NGUYEN: Thank you so much, Chris?

     >> CHRIS WILSON: The partnerships, like the -- like we just talked about the AI technologies today, but if you think like AI -- like -- (indiscernible) work and then they have to make a solution, just solution for the handicapped people, for the disease and everything. So, in that sense, the cutting-edge technology is going faster -- but on the other hand, we have to get the effort to make it to the solution. More carefully, before the climate changes and those kinds of things. So then, so my point is like the AI right now is more like a developing state and still -- even though we have more of 60 years of history with the development of the study of accessibility and inclusion. So now maybe we -- it's time to think about the diffusion of the technology and the beginning of the -- (indiscernible) that 1 billion people with the handicap waiting for the technologies. And before the AI comes, I don't think we had the technology, the outcome of the technology or the device to support that these people -- (indiscernible)

      So, for the diffusion of the technology then, like even though the Microsoft and the other companies are watching and waiting, the technology but on the other hand, everybody has to be eager in the process of the technology is not only the company's fault, it's not only the government, you have to get the -- (indiscernible) or get together to make that solution. I used to work for the Japanese government before I came to the OECD. I know some of them. The company will pay the companies to hire the disabled people. That's a traditional way we continue to do this.

     But on the other hand, we have to find a different way, a new way to get better employment for the handicapped people. It's not just the government effort, I mean every country, even Japan, has a constraint on finance. So, they'll have to create a better way to do the collective effort to empower our people including us, even -- (indiscernible) the great people that takes care of the handicapped people. And I just found that they are not in many cases less familiar with the -- so maybe if one of us can help understand the technology, then -- a good use of the technology. Then they can see that -- (indiscernible) maybe they'll take that. That's what I hope? Okay?

     >> Sorry. Let me take what -- (indiscernible) so I think that we understand the platform, the services, on products in Latin America. And we take quite seriously our role in terms of the summation of information and explaining all of the benefits and the -- (indiscernible) obviously coming from the equality. Governments and to explain and to work on -- and the policies of this. Our focus is going to be -- (indiscernible) one is artificial intelligence. We need cloud to provide services. And I think we have a little space to work on this, the creation of awareness is crucial, not only in -- (indiscernible) and I take quite a -- I think it's what -- quite interesting to be -- (indiscernible) to hear your experiences.

     >> CAROLYN NGUYEN: Technology to be able to build on that trust. But I'm hearing it's also important to create that awareness with the people who are so they do know that the technologies are out there as well that can enrich their lives as well as from government for the need to provide incentive. So, awareness education. And that's critical in making it sure that there's efficient diffusion of the technology to enable the 1 billion people, the majority of which are in developing countries. Through public/private partnerships, through shared responsibility with all of the stakeholders at the table. And also to make sure that there can be a holistic enabling environment to incentivize everyone to focus on this issue. And in that wholistic enabling environment, we need to be thinking about the social culture of impact and value, the economic impact and value as well as the development of new technologies and what's the governance that can bring all of this together as well. Thank you so much to all of our speakers for sharing your perspectives and insights.