 So the institute was established with a goal to become a global leader in certain areas of applied computer computing research and the key word here is applied. We wanted to do something that would have an impact on the society and the residents and citizens of Qatar and the region. The way we would do this is by conducting innovative research in, you know, across multiple disciplines in this applied research. And so it had to go along with the priorities of the state of Qatar, of course, which is enhancing the quality of life for the citizens and society, enabling great scientific discoveries. And that's been happening quite a lot here in Qatar. You're quite proud of the progress that's happened so far. And also, enabling the local businesses to compete globally. You want to give them that edge so they can actually step out of the boundaries of the country. The institute was founded in 2010, we're at about 125 employees right now. And the nice thing about the institute is that it has a very big mixture of scientists, post-docs, program managers, software engineers. But the better thing is that we have a rich mixture between academia and industry. Now, if we're doing research just for research purpose, you know, it's good to have the academia. But if you want to go further out, if you want to make products, commercialize, it's always great to have that industry perspective in there as well. And also, a great perspective is the diversity that we have. Over 25 different nationalities that really, you know, lens it in together. And we have a very aggressive plan to triple the number by 2018. So we're aggressively hiring and looking for the right talents. As mentioned, we're not doing research for the sake of research, right? So the areas of research we do are very carefully selected. There has to be a specific problem that needs to be addressed. And so, at this point, we have these six different areas of research. The largest one is the Arabic language technologies. It just falls, you know, correctly that we are in an Arabic-speaking country. The Arabic language is lagging behind a lot in the computing sense. And so we really need to add this energy into it. Computational Science and Engineering, I'm not going to say much because I don't understand any of it. It's about genomes and things like that. So just wait for my big rate. Cybersecurity, that is very important. As you know, Qatar is actually a, it's got a big bullseye on it, you know, from quite a few countries in the region and elsewhere. And so cyber security, there are a lot of attacks and threats that happen. And cyber security was identified as one of the three big bets or challenges for Qatar, you have the water threat, energy security and cyber security. So we were tasked with leading this effort, working with the Ministry of Interior and other stakeholders here. Data analytics is also a very big research area that we have dealing with big data. I'll talk a little bit more about that. Distributed systems, cloud computing, video processing, so converting to these 3D video and projects like that. And finally, social computing innovations. That's a place where we've really put our mark in the world. We've got a lot of projects here dealing with, specifically, with human disaster areas, recovery and crisis management. So I'll talk about that as well. And please, if you have any questions, just go ahead and interrupt. Yeah, I think it's fine, we don't have to wait till then. And so we're quite proud of the achievements we've done so far. In the past three or four years now, we've got quite a lot of citations, publications and we have filed over 71 patents in the US and in Europe. And five of these have already been granted. And there's across the six different areas of research. And we've actually spun off one company. So this was a nice achievement. It's called DataTamer in the US. We've got several technologies that I think we've gone past it for. Several technologies that are currently licensed and being used by businesses. And quite a lot of different software has been deployed. Not that we're on form, but we've had quite a bit of mention in a lot of the top media around the world. This is just some of them. So it's really nice to see the name of an institute that's local, that's here, being recognized worldwide. So moving on, we'll talk about the Arabic language technology, since this is especially about Arabic content. And I'm glad I went first, I was telling Khaled, because content is really boring compared to robotics. So at least he don't have to compare me to what he. In the Arabic language technologies, we've actually done quite a bit of work here with developed software, we've actually filed patents. And just to give an idea, in the Arabic language technologies team, we work in two tracks in parallel. One is the technology, and the other is the content. So the technology is what's going to enable the consumption of the content. Without that, really not too much in terms of the language. So with regards to technology, the core of using a language by computer is natural language processing. This is the foundation, the cornerstone. If the computer cannot process the language, it really cannot do much with it. And what you see here are all technologies. These are not products, they're not solutions. Once we have these technologies, it's the job of the entrepreneurs to actually figure out what to do with this. How can you combine these technologies into a solution or a product? We've done some of that, but we are a research institute in the end, and so this is really not our end rule. With natural language processing, it's things like, how do you split up a word? How does the computer know this is a noun, a verb, an adjective? An example of a technology that can become a product, if you want to do text-to-speech so that the machine can actually pronounce the words. In Arabic, you have the diacritics or the short vowels. Without that, it's almost impossible for the computer to disambiguate meaning of the word, what does that mean? So for example, Al-Alam, Al-Lima, Al-Lama, it's the same word, but the vowels are different. So you really need to have these diacritics. To have these diacritics, you need to know what the word is about with the context, so we built a diacritizer. You just put in the text, it will actually put these vowels for you. So this is a technology that can be used into a solution which is the text-to-speech search. If you have the information, you have to have a way of getting to it, right? So information retrieval, this is where we built our own search engine. When QCI was first established, one of the goals was build a super search engine that would put Google to shame and make. But that's not really a research program in a project. And what was decided instead was to focus on a smaller focus, have a smaller focus to work on the Arabic language. So we have an enterprise search engine that you can deploy in your company's database on your website that will have Arabic specific features. So when you search for a word, it actually breaks it down to the root. And then it expands your search results. One example I always use is that when, I haven't done that in a while, so I'm not sure what the results are. But previously, if you search in English for the value of the kings, you will get all the information you need about the value of the kings in Egypt, like certain permits. But if you search in Arabic, what do you look? You get zero, nothing, not a single hit that is correct. You get what is it out in TV shows, basically, TV series. So nothing that has to do with what you're really looking for. And this is where it's important for the engine to understand what you're looking for and try to find it. Machine translation is also extremely important. Yes, sir. Is that a problem of content that it wouldn't be searching? It's both. Well, if the content is not there, you're not going to get it. But if it is there but not tagged properly or you're not searching for the exact words that you, and the search engine needs to understand the context to try and find things for you in context. So for example, we have a video search right now that will find what you're looking for. But not when you, for example, if you search for Daesh, which is ISIS, it will find for you ISIS, it will find the Islamic State, it will find ISIL. So anything that's related to it, similarly, if you search for a person's name, Mahmoud Abab, Mahmoud Abbas, and find Umaz and whatever relates to that. So it's both. The content has to be there, but it also has to be searchable that you can find it. Machine translation also uses a lot of the technology. The democratization is very important for that. If you have that, it will help with the outcome of it. We built our own engine as well. This is the machine translation. We've entered it in a couple of contexts, competitions. And we've gotten first and second place. So we're also pretty happy with the results on that. Audio transcription is extremely important for many reasons. I'll talk about that in more detail. But this is basically the speech to text. If you have a video or an audio, and you want to see the closed captions on the screen, so we do that for Arabic. Optical character recognition, scanning a document, you want to digitize it. There are a lot of softwares out there. They don't all work very well. And there are specific challenges. So if you go back to the old manuscripts, the fonts were different. They're handwriting. Letters were not always connected properly. So there are a lot of challenges. So we're doing some research on this. How can we improve that process? We do a tweet analysis. This is really fun. Actually, we've been collecting every Arabic tweet for the last three years. It's about the tune of, I think, 10 million tweets a day. And then we do our own analysis on it. You can understand the rudimentary sentiment analysis at this point. But Arabic is really difficult to try and figure out if it's positive or negative. But we do that. We do a lot of, I'll actually tweet moogers, which is one of our products right now. And there was just a recent research done by one of our team members about the propensity just from analyzing tweets, understanding if a person is pro-ISIS or anti-ISIS, or if they're going to be pro or anti. And this was done and actually picked up by quite a few news organizations. It was very interesting. Just from tweets, from the hashtags, you can get a very good idea of where somebody's ideology is going. Yes, as the events happen. Exactly. And this is the case with tweets that always have to be related to an event or an article or something that goes with it. And then we have a lot of work on educational applications, and I'll show those as well. So on the content side, we've done quite a bit of work. It's a little bit focused, though. So we've done work on Arabic Wikipedia. We're working right now on enriching the Arabic medical content. We have a social interaction platform. We haven't launched it yet, but it's ready. And we are working on an Arabic book reference. So again, I'll talk about those. Absolutely. Most of the stuff we do is open source. And we actually, we love to collaborate, especially if it's a local initiative. So you have all the contact information or whatever we're more than happy to. So since we're talking about content, I mean, the first question is, do we have a problem with? Everyone's saying there is not enough content. So if you just put the word, as a mental methodology, the prices of Arabic content will be 902,000 hits. Obviously, this is a combination of these words. So it's not all about the prices. But if you search separately the problem of Arabic content, lack of Arabic content, you'll get thousands of hits for each one of these. So at least people are talking about it. So if it's not a problem, there's something happening. As is commonly known by a lot of, the numbers being used is that the Arabic content is 1 to 3% of the entire content. And that is a very, very small number, given that Arabic is the fifth most vocal language in the world. And so looking at 3% and 3% is probably on the higher end. And to make matters worse is that out of this 3%, 75% is what we would consider not valuable content. I mean, valuable is a relative term for someone. A movie is valuable. But if you look at the educational value, the research value, the scientific value, 75% is in the form of entertainment and forums. And to make it even worse than that, 80% is actually recycled and not original. So the original Arabic content that's valuable is minuscule. No, not necessarily. Recycled as in the same content repeated over and over again. So Arabic is all Arabic, yeah. We don't differentiate from the source, whether it's original versus translated, as long as it's Arabic. So what's the reason? Is it lack of people that use the Arabic language? Not really. I mean, the figures here show you that Arabic is the number of Arabic users on the internet are the fourth in the world, which still only makes up 4.8% of the internet users. But still, it's from the ranking perspective. It's not bad. But if you look at, so what's the other reason? Is it possible that they don't have access to the internet time? Actually, internet penetration, yes, it's lower than the world average, but not considerably lower. So 36% of Arabic speakers have access to the internet versus 39% for the world average. So maybe they just don't care about Arabic language that they don't want it. But from a survey that was done about what language would you prefer the content in? 60% actually, more than in Arabic, for the Arabic speakers. 33% in English, 60% in French. So it's there. They want it. We were just talking that there is a problem. There are a lot of problems about the lack of the Arabic content. But basically, that just transforms to a lot of opportunities. People want it. They just want it in the right form and the right way. So all of the, exactly, all of the Arabic speakers who use the internet. But now the survey, I honestly don't know what the sample is. It's from the Arab world online. I think the previous slide comes in a little bit. These are the speakers, the number of speakers, yes. Yeah, on the internet. But how many people were actually used in the sample last? So what are the challenges here? I mean, why do people, why is there this problem? So when, again, we're serving the same report as to why, what the challenges against using the internet. The first two, accessibility, connectivity, and cost. This is more of a global issue. It's something that is everywhere. But then the next one is lack of content in my language. So 41% stated that the lack of content in their language prohibits them from using the internet. Now, we're also talking about this with Sarah. When I was back at Microsoft, I was in charge of language planning, deciding which languages do we localize Windows and Office into. So when I left, we were doing 115 languages. But deciding which language, the number one priority was for monolingual users. Basically, if you know this country, majority don't speak English or French, or one of the major languages. Then you really have to provide them with the software in their own language. Otherwise, they cannot use it. If it's a preferential thing, it's a lower priority. So in this case, there are a lot of other countries where you have monolingual users, whether it's in Saudi, Kuwait, in Egypt. And they really require that in their own language. But also, it's not that bad. It's not all dark and gloomy. If we look at Wikipedia as an indication of where things are going, in 2012, there were 318,000 articles. Today, it's almost doubled at 625,000. So that's a pretty good increase over three years. It's acceptable. So at least we're moving in the right direction. But at the same time, when you look at the ranking of Arabic, Arabic is ranked 16. So yes, it comes after Chinese, Portuguese, and Japanese. That makes sense. But then when you look at languages like Sebuano and Wariwari, Wariwari is the fifth language in the Philippines. So it's not even a primary language. So for that language, they have double the Arabic articles. In Dutch, 20 million speakers, they have almost 2 million articles. We have 600,000. So there's a lot of work to do. And when you take a deeper, you'll see that the reason is the number of editors is considerably smaller for Arabic. The creation of articles is smaller. We have a lot of translations happening. But we need to put a lot more effort into this. Yes and no. Because even for other languages, there's really not when we talk about Wikipedia, the biggest incentive for Wikipedia is self-critification. You put up an article, and that's there for you. So for, in this case, I don't think incentive is. But in general, yes, incentive is a big problem. You have to send them to create or to do something. So what we did here at QCRI is we have this initiative called Ifran. Ifran means enriched in Arabic. So we want to enrich the Arabic content. So this is relevant. It includes all of the different projects. So specifically, with Wikipedia, the first thing we did was did professional translation with 10,000 articles. This was more like, OK, let's get something going. Let's move people, let's explain them to do some work. Then we used working with the Arabic and Canadian community. We used bot to create 100,000 articles in the geography domain. And bot is a software that actually takes the English article, translates it into Arabic, but it's a seed article. So it's very small. It just takes the important information, puts it there, for others to go in and add to the article. So whereas it took us about a year to do 10,000 articles, professionally translated, very expensive, it took a fraction of the cost and about six months to do 100,000 articles. Sorry. The bot, which is the software that just picks up the information it was in. You don't like it? No, no. The Arabic and Canadian community. So there is a lot of these bots. And at first, we were reluctant because we don't want to just add articles for the sake of saying the number went up. But when we compared with other languages, actually Dutch that has 2 million, I think 90% of the articles were bot created and not even created. So there's a lot of this. So at least we put 100,000 articles about cities, geographical locations. If somebody doesn't speak English, they go put a name of a village in the US somewhere. At least now they know where it is. The company, the population, and so on. Finally, we also created an online platform for translating articles from English to Arabic. If you go into a Canadian community and try to translate an article, it is extremely cumbersome. You have all of these formatting tags to take care of. If you miss one out, something goes wrong. What we do here is you just log in. You type the name of the article you want to translate. It automatically extracts all of the text that needs to be translated. You put it in a nice, editable editing environment. You can even do a Google translate as an initial step. And then you have to post-edit it to make sure that it is correct. And then you just push publish. It's out on your account. So it makes it so much easier than actually having to do it on the Wikipedia site. For free? For free. If you want a baby, that's fine. We can talk about that. For the health content, this is in general, content is bad for our baby. Health content is really, really, really bad. We've just started getting some websites that have some credible content. But before, if you ask any medical question, it's more about forums. My grandma does this. My aunt does that. So it's really hard to get any credible information. So we went to Mayo Clinic. Mayo Clinic is an authority on this. We partnered with the National Library. We licensed their entire medical library. And we are now translating it into Arabic. It's currently being translated. And will be made available through the National Library. So at this point, will be made available to the residents of Qatar, since each resident has access to it. So our license is for Qatar only. But this is another place where I talk about opportunities. So now that we have this Arabic content, our door is being knocked every day by ministries, by hospitals, from all over the region who want this Arabic content. So now we will start licensing it out to them. So this way we can recover the cost of the translation. But it all ties in together very well because by doing the translation, we have the translation memory. We can improve our machine translation engine. We can, it's all part of the research as well. We're working on the Arabic Book Reference Project, which is somewhat on hold right now, but basically we're going to put the web to get the official information about all Arabic books that have been published. The book names, the cover image, the author and all of that. And we want to make that available so that you can find, you know, any book you want to know about, you can find the information. And then the next step would be if it's available in digital format, you can click and actually see it. Obviously this is for the public domain books, but then we can do something for the cover articles as well. The interactive social platform, I'm not going to do the demo, I took some screenshots just to show it. Basically I wanted to find a venue for people who are interested in Arabic content to talk about it. So it's more like a Facebook, but for, you know, Arabic content, people interested in Arabic content. You log in, you have the carousel and profiles of the different people. You can customize it as you want. You have your timeline, your profile. You can post stuff and people will like it. They can share it. We have a word system, a chat functionality. So it's an entire social interactive platform that you can work on. And we hope to launch it fairly soon. The next step would be for that to close the loop. At this point, it is integrated in the sense that you can opt in when you first log in to say whatever I say here gets posted on my Facebook over by Twitter account. Yes, in that sense it is. We haven't launched it yet because, okay, we built it, but we don't really have a team to run it. Right, this requires an entire infrastructure. And so we're working with the World Organization for the Renaissance of Arabic Language, which is a new center at QS, and they're responsible for enriching the Arabic content as well. So we work very closely with them. They could fall very nicely into their lab. And this is what we talk about, you know, creating technology that can be in this content. So if we switch over to technologies, you know, a lot of the challenges currently with the modern Arabic language is, you know, these things here. So dialects, this first sentence was taken from the web. I wouldn't, the only word in here that is actually Arabic is the name Muhammad, right? So Muhammad Darbatov, who built that with the reckless guitar. Muhammad was hit by an automobile and was sent straight to the hospital. And so, you know, the foreign, you know, and many foreign words like, I think North African, Moroccan or, decoration, you have, you know, decorating the text using characters that are not in Arabic. So here you have Persian, Urdu, elongation where you wanna put emphasis so you put a lot of letters in the same letter you repeated. Or pointing in your terms, this is such, such mean, subtle, you know, truth. And so, that's not an existing word. Arabian, which is the Arabic written in Latin characters. So, while we, we don't wanna encourage people to do this, right, but it happened, let me mean, this is what people do on the social media and all that. And there's a lot of good content there that we don't want to use. So this is part of what our natural language processing does. It actually normalizes that text. So it looks at the decorated text, removes all of that, converts it to proper Arabic characters. Then it goes to step next, which is converting the dialect into more extended Arabic. So we're doing a lot of work in that area, similarly for the Arabic. It's a very tough job, but it's very rewarding to see that happen. So I don't wanna take too much time, but with the natural language processing, there's a lot that needs to be done. This is our entire stack of things that we work on to get us into the solutions that we wanna have. But I wanna talk about this product specifically, because it's very important, the QCRI Advanced Transcription System, CATS. This is the speech-to-text. So everybody knows what the closed captions are, right, on the TV, you get the subtitles. That exists for many providers for English, but for Arabic, very few actually companies do that, and the quality is not always at its best. In our case, we have our own state-of-the-art system that provides a guarantee of 85% accuracy, and it's currently used by Al Jazeera. So if you go to the website, you look at their videos, if you hit on the closed caption, you will get those, and that's using our system. And on average, the quality is about 90%, it's only a 10% for their rate. It's great because it makes the content searchable, like, I know I watched a video and there was a specific part that I won't. The best case is if I know the name of the video, I can find the video and then I have to sit for an hour until I find what I'm looking for. If I don't know the name of the video, pretty much you're out of luck. In this case, when it's been transcribed, you just search for the word, it takes you to the video, you click on the word, it takes you exactly to the location of the form. And our next version allows for context search, so basically you don't have to search for the exact word, you just put the idea that you're looking for and it'll find for you that segment. And now we have actually dialect recognition, we recognize the top five dialects, so it will tell you right away if it's Egyptian, Levantine, Moroccan, Iraqi, or Gulf, sorry, Egyptian or Gulf. Unfortunately, this one is an offline solution, which means it's not for instantaneous transcription. However, we are working on it, we are our prototype right now, so hopefully within the year, we will have the instantaneous one. And this is an example of how you can productize technology, so I'm next week going to meet with CNN, Sky News, and NBC to market this, because this saves them a lot of money and a lot of time. At Jazeera, they used to do this completely manually for the shows that they wanted. Excuse me, yes. I'm just thinking, suppose we have an English review on you. Can we have some translation thing, which you know the guy speaking English, and then you have the subtitles coming in Arabic, or whatever he's speaking. Absolutely, thanks for bringing that up, because we do provide that already. So if the guy is speaking Arabic, you have the option. He's speaking some other language, but the process is different. The one we implemented is he's speaking Arabic, you can switch to English, so the text comes out English, but it's easily done in the other way around. The reason we didn't do the other way is, since this is an Arabic language technology, is we only developed the technology for Arabic speech recognition. To do the English is a different language model that we weren't interested in, because it already exists, and we can acquire it from somewhere else. But yeah, that's definitely something. So we actually were part of the BBC News Hack, it's a competition back in December, and we're invited to it very late, so the team just went there, and in two days, we won actually Best in Show, basically taking an Arabic BBC page, translating everything into English, taking the Arabic video, transcribing it into Arabic, translating it into English, and then synthesizing the voice into English. So now we went from a completely Arabic page to a completely English page, and that was an amazing achievement in two days, so it was actually pretty good. So we have a lot of educational apps. The app was built by interns during our summer program. It's just teaching Arabic for non-native speakers, very basic though. Medaille Group was an actual wheel that was developed by a further foundation international, where you get the Arabic letter and then the corresponding English letter, just so you can read your name, you know. So we took that and we actually developed it into an app, again by one of our interns. The last one, the Jalees, is there are not that many e-book readers in Arabic, so Kindle, for example, doesn't support Arabic. iBook supports Arabic, but not fully, not natively. If you open an Arabic book, it goes left to right. So we built one on the iOS, but then the Supreme Education Council, they launched their e-bag initiative, which is replacing the traditional textbooks with a tablet, running on Windows. So we actually built a Windows app for them. There were five companies bidding for it. We managed to meet all the requirements, so we got this. It does e-pop free and PDF. Interactive content and multimedia, so it's really embedded in the videos and all of that. Works on iOS and Windows 8. And it's right now being used by 40,000 students in the independent schools. So that's what I mean by native support. Our reader actually supports Arabic natively, so it understands all of that. In fact, we also have a patent for language detection. If the book does not have the metadata telling the reader, this is an Arabic book, so it opens it from right to left. It actually goes in. If it doesn't find the language ID, it goes in, grabs text, compares it to the code page. If it's Arabic, it makes it right to left. So it handles everything from perfectly. In fact, I have it here. I can show it. So it's currently in use here in Qatar. And again, we are a research institute, so now we're depending on, in order to take this and see what can they do with it. Can they market it into a bigger solution, like part integrated into the learning management system that is used by schools. So right now we are in discussion to have it used in seven different Arab countries. Kuwait, Egypt, Jordan, Saudi, UAE. Tweet mogas, this is the red box. So very nice app. This is what I was talking about, analyzing the tweets. So we did the engine and all of this, but then we licensed it to a company in Egypt called Better IT that created this interface and the customer facing approach. So basically, you have all of the tweets coming up here. You have a word map and then you can select one politics Syria and brings up all of the tweets that deal with this. And the word Syria doesn't have to be mentioned. It actually gets it from the context. You can filter for text tweets. You can filter for image, for videos and all this. So it's really a great application. Now, this was used by Al Jazeera when during the Obama-Romney debate, when they had that debate on the bottom of their site this, all the data given was using tweet mogas based on the tweets, basically just understands what, you know, how many times this was mentioned, positive, negative, so on. We have a project called the meeting translator. Again, that's, if you're in a meeting, people don't speak the same language. You should be able to understand the other person. So whether it's lectures, presentations, so it's basically an interpretation. And we have a prototype already, so we're just improving on it. Very quickly, other projects outside of the architecture technologies. Nadeef is the data analytics. Nadeef means clean because it cleans up the big data. So here we have a seven-year joint project with MIT, the Community Science, Artificial Intelligence Lab, working on this. And we also have a two-year contract with Boeing. Boeing gets a huge amount of data from their base. Their efforts are continuously streaming data about the performance of their specific parts. What we do is we curate all of this data, clean it up, find out what the trends are. So one of the things is predicting the, when a specific part is going to fail. And that's been very accurate so far. And so, Boeing is using it for two years now and then we'll see what else we'll add to it. A, they're Artificial Intelligence for Disaster Response. This was actually used in a lot of the disasters, whether it's the Cyclone in Manwatu, the Typhoon in the Philippines, the Earthquake in Haiti. It is an application that links people who need something to those who can provide it and to allow to, the next one, MicroMappers, uses that information to put it in a map where there is damage, where there is need for food, need for shelter, or somebody has something they can offer and that's used to link people together. Fast is used by, Fast and Ears, both are used by Al Jazeera. Fast uses tweets about a specific article to predict the life cycle of that article. So we say, okay, this article is going to get 4,000 views. It has now received 2,000 views, so it's halfway through. And Al Jazeera would decide, you know what, this is an important topic for me, I really need it so they'll put more promotion at it or take it down if they don't want it. And that's been, our accuracy rate has been extremely high in the high 90% in terms of what we predicted. Ears is when an author writes an article, they have always the related articles linked in the bottom. The articles that get related are only articles that the author knows about or that they've actually authored. So it's really very restrictive. What this does is it actually recommends articles to the author. So when the author says, okay, give me recommendations. Just from the context, it'll put the list of articles and then the author can choose those. Verily, when you have a disaster, you have a flood of images that come out. A lot of it is fake. They have this big shark coming up to eat the diver coming out of the helicopter. So all of these things that come out, Verily is a crowdsourced platform that allows people to say this is false or genuine. And then the last two are the video. And this is amazing. So 3D videos haven't really caught on much except in the movie theaters. And it's a chin and egg problem. The making them is expensive. And so manufacturers are not making devices that are cheaper. And so it's never ending. So what we did here is we actually convert 2D videos into 3D. So you don't have to spend the money when you create a 3D video. And when we did a sample test, the results were that people actually preferred the converted one to the original one. They said it was more comfortable. It felt more real for them. And it set the fraction of the cost of creating the 3D. And then finally, it's the 3D video streaming. How we can get now 3D videos on your smartphone. Pretty much everything is developed in house, yes. What? See, OK, that's where entrepreneurs come in. So we are a research institute. When we did CATS, the transcription system, and we offered it to Al Jazeera. Other were partners with us. We actually did not even have a mechanism to get paid. We paid. We never get paid. So we had to send them to see, OK. So now that we did this, Al Jazeera, well, their partners with it, which is fine. But if we go now to CNN and say, do you want to use it? The first question is support. You offer 24-7. We offer support from 9 to 5 if Mejh Ahmed and Ahmad Rahman are available. So that's why now we're looking at the possibility of spinning off companies. So let's spin off a company for CATS. Because CNN may come back and say, I love this idea, but I need you to do this and this and that. So that means we need to hire one more person. QF has a hiring freeze. OK, if it's not a hiring freeze, then you have to go through a committee to find someone. And all the bureaucracy doesn't take forever. But if you actually create a company, then that not gets much simpler. You have the freedom to do what you want. And to what extent is that available? No, it's not. Actually, the initial approach is that everything will go open source. But now we're kind of re-evaluating. If we're going commercializing, then this will be different. So for example, the reader, the evil reader, it's licensed to the Superimmunication Council. We are going to release it at the Windows Store for free. We'll do a little bit of differentiation. Maybe put some restriction. But because we want to still license it now to the administration, the UAE in Egypt, and if you make it available for free, then it's gone.