 So yeah, so first of all, I think like before introduction, this is nothing to like nothing related to Amazon, first of all, now, so I was working with Buckman Center Howard University till last August 16 and like so we tried to explore all the projects that Buckman was working on the world was working on. So usually my Jigsaw project was not related to machine learning or anything so but we had a very short discussions every day and we like tend to find out what exactly is going around our industry. And so there would be nothing technical in this so like you can wait for at least like 10 to 15 minutes before your lunch. Yeah, so I'll discuss some of the good projects that are like I came across or my colleagues came across and usually after working hours, we don't have the energy to work like board something. So we usually discuss on like what are the exactly impacts of AI. Even if you Google right now that what exactly is doing us to us after the 10 first things you will get they will always discuss about like what will I do rather than what it has already done. So here like lots of most of the projects will be exactly on what already is done to us or something on proven grounds rather than theoretical ones. Yeah, so mostly the first one will be discussing on what exactly the present in entrepreneurship scenarios and what exactly verticals like companies are working on. And then we will have a short notes on like each vertical of our world and the impact of AI on that. Let's start with the industry. So this is the like a small survey as well as like big research done by CB Insights and they tell what exactly AI scenarios improved. So these are mergers and acquisitions that has happened in from like 2011 to 16. And like we believed at the end of 15 that there will be a probably small break or small winter in AI, but that has not actually happened. So it was because of the like good projects that had Google and Facebook like Facebook and that came up and through all the awareness to the people. And that's if you see the 16 2016 it is always increasing. So a lot of them are good companies, major companies. They acquired very small startups and even the investment scenario has changed. So these are the deals. If you see like in the quarter three and quarter four of 16, sorry, second and third, you will see that like lots of money has been invested in AI and like the number of startups that started with AI has been like a lot. So this is the overall scenario from 12 to 16. So I came across a small machine intelligence landscape. They researched from 2012 till now and they like crafted out the entire scenario. So this was what started in 2012. You will not see any of the open source projects that helped the people to like develop more on AI. After two years, this was the scenario in 2015. This was like MI 2.0. You will see that the small difference that in 2012, there was nothing, not more than four to five companies working on each vertical. And it was like very small, like probably five to seven people working in each company. This was the scenario in 2015 where like lots of companies grew up, but still the companies were small. And like a lot of good projects came into research. This is what last scenario is. That was in November 2016. You will see that like lot of the projects from Google, Facebook have came and mostly were open source. Lots of open source tool came up, open AI started like some of the research projects. And they also like threw one more application into the public that was Universe. So yeah, let's start with what exactly was done in transportation sector. So you might already have like heard about autonomous cars, semi autonomous cars. And the main impact of this was that government actually trusted people, trusted the companies that actually was developing the semi autonomous cars. And they have like in San Francisco, they've already approved the fully autonomous cars. So still now only semi are allowed. And yeah, actually at the end of 2017 probably will have fully autonomous cars on road also. Yeah, of course, the lesser probability of accidents like human committing an error is like, but if you see the probability is like quite high likely than what like autonomous cars will do. And yeah, there are still loopholes in if you see the picture, you will know what exactly I meant. There are lots of algorithms that are still to be solved. Transportation tool. So like we took a step ahead from cars, autonomous cars. We went to drones and from drones like merging drones and cars. We like Dubai, if you might have followed up on this recent news, you are meant to know that Dubai is already developing a small POC and they've already showcased that on one of the recent events. So they will have a taxi, small kind of taxi that will be like fully autonomous, probably at the end of 17 on road. Yeah, employment. So like there have been a lot of talks that AI will like snatch a lot of jobs. Basically, if you'll see again, if you research, lots of the research will always have will that AI will take your jobs away and there is no proven grounds that it will actually do. In fact, there are a lot of debates that I will create more jobs. So there is only one paper till now like published by organization for economic cooperation and development, which says that after the 15 countries that is in the group, 9% jobs have already been reduced on for UK, the scenario is that 10% jobs have been reduced in effect of the AI. So yeah, and there is no other paper that I am able to find on. So of course, productivity have been increased. We'll see how and how like where exactly it was affected. And yeah, of course, the like if you see AI after the entire 100% of the people only probably 20 to 25% of the people will be able to afford the autonomous cars or the all of the technologies that will be able. We'll see AI affecting but so it is anticipated and like nothing, nothing on paper, nothing on survey that it will widen the gap we have between the rich and poor as well as the educated and non-educated. So probably some less educated people might take some more time to get adapted to the new technologies. This was the scenario for the main occupation categories that employment share for exactly what changed between 2002 and 2014. So the last decade actually was the main like AI impact decade. And if you see that the minimum the medium main medium routine one was it did the entire jobs reduced to around minus 8.25%. So that is a too much for like so the main inference is that maybe the top level employees might not be affected but the middle class and the lower wages. Might be get affected the ones who work in the industry or factories and all. Yeah, so like lots of lots of things in robots like so the first one was that they they become cheaper. China actually like had a lot of like four to five big companies working on pushing the robots into each home. One of them is UbiTech where they developed like life-size robots using Amazon Alexa. And they can do like small tasks like like listing the shopping list or something like that. But they became so cheaper that it could be like already into the homes around in a month. Also, there are a lot of tasks that are like dangerous and like hazardous probably so one of them is the nuclear waste cleaning. A Japanese already used that for around two to three years now and there was a recent news also that decaying has also occurred in those robots. And so even so the recent developments has also said that they are developing such robots that cannot be decayed by the radiations. And also there are a lot of tasks like simple ones like painting high walls buildings where like people have died a lot of times before. And so like NTU here has also developed a small prototype which can like easily paint the entire building and yeah. So also there have been like we have seen in a lot of the movies that maybe after some time we'll be able to control robots using brains and all. So there is small prototype of that too. I'll skip this one. Yeah. So the prototype is here. It was recently developed by MIT where you will be able to tell the robot that what you are doing is actually wrong. And we'll be able to train the robots very like nicely and as fast as possible. Okay. So education mostly like a lot of people have told that you will be able to replace the entire teaching scenario. But actually most of the people have said that it will not be possible as far as the face to face integration is concerned. So like you will be able to your teacher will be able to understand you more than a robot is what they've claimed. But and so like till now mostly teaching assistants have been developed till now. So there was one talk on TEDx by Ashok Goyal that he developed a small assistant and then so like there are four tears. And the fifth one was called jail Watson usually like entirely built using IBM Watson and they actually didn't tell the students that it was an actual robot till the end of the entire course. So this is a small conversation and even you will be not be able to make what exactly so one of them is by the teaching assistant and the other one is not. So the left one is so yeah if you calculate the long term learning and like teaching improvements. So you'll be able to find out that probably will take more time around a decade. So I don't have the data for exactly for that to replace the interest to teaching a scenario. Also again like Dubai Dubai again has like a small prototype coming up in end of the 17 which will be assisting teacher on like teaching a repetitive task of every day. So a lot of public safety and security robots are like being deployed everywhere where you like so the entire data is being used for like predictive policing which where the like the robot will tell you exactly what can go wrong. Where before 24 hours or online algorithms also that something could go wrong probably and the accuracy is more than 90% of which has been achieved till now. Yeah. And like so night scope security robots in mountain California is actually in production right now and it has like helped a lot in a lot of cases where to find out like any sexual predators or any of the thieves. Like just on time also again in news they have said that in 2018 will be having robots assisting in the Olympics also. So hopefully we'll be able to see a lot. Yeah. This is a small prototype where actually the entire so like there are drones which takes the entire heat map of the area and it will tell you where exactly like what kind of emotions or like anger or anything is there in the general public and it will tell you that these areas are prone to like crimes. Yeah. Healthcare there have been like almost I guess three to four like breakthroughs. One of them is like finding the use of the existing drugs. So whatever we have already built can it be used for any other diseases also. So there was a request and farmer farmers which have worked on around like 15 rare diseases which like we exactly don't know what exactly the solution of those diseases could be. But they have been mapping by like seeing the cells more clearly and then mapping them to the drugs that have been built also Stanford research recently sport cancer more accurately than what humans could have done till now. And they have like so by identifying 10,000 individual traits they find out that if this cells are like developed like by this way and this could be the results. Of course you might have already seen this video. So it's also fun to just match the keyboard. The neural net tries to return something coherent from any input they give it. So this is one of the Google AI experiments. One of the like individual has developed a small duet where you like play some music on piano and it will like repeat with you. Not exactly repeat it will play a duet with you. So just I'll play again and you get the idea fun to just match the keyboard. The neural net tries to return something coherent from any input they give it. Again, two of the another two main like projects that came up one was dancing lights. So there are a lot of street plays and there are a lot of stage performance going on. But you need to have like manual people working for you who can move the lights for you. But there is a prototype where the according to the artist moments and so this isn't dancing the lights will dance. And one of the another like great tool by Stanford is to text scene. I'll just give a small demo of what exactly it means. It's like you write the text and it will convert the exit text to this any of the scenes that you tell. So if I say generate a room with a desk and a lamp it will create a room with a lamp. The good thing about this project is that if you repeat the same like commands again and again it will like it had a lot of database very big database which will create like different rooms for you. Yeah, in a good culture actually we are still missing a lot. So if you see like we are able to like detect the crop diseases but it is actually only on the basis of the images that we are getting of the crops. There has been there have been a lot of like Android applications where like the farmer can click the photos and we'll know what can go wrong what diseases it could have after probably a week or two or what exit it has right now also. But still so though we are able to like attain 99 around 99.35% accuracy like we are still a lot like back in terms of number of species will be able to find out. Also, so before three to four months there have been a lot of pilots going on where the drones could be like over around the entire area and can tell you what exactly the crop yields will be. So it was actually like very breakthrough moment because through the images it is like really hard to tell what exactly like crop yields will be and even that accuracy like we have attained till 97.5. Also, you might have seen like it is very famous video of a Japanese farmer son working on his farm where like the cucumbers come in and he has developed an entire hardware where it will sort out automatically according to the shape and all. The exact cucumber type like is going what is like what he receives. And yeah so the most important one is the social so if you see the sentiment of the people like it has it have been like like very very lot of ethical issues that what like what exactly is doing to us and but still so more than data it is just a debate so you will find it like everywhere. You will probably have like short conversation with your friends on the same. So yeah there are three most important ethical issues like what humans do can do ethically to build the AI what AI can do ethically to support us and what will be the interaction the same that this is a vicious circle and what will be the diameter of it so how this will happen. Also there have been a lot of steps by like great people Bill Gates Elon Musk that they have signed an open letter that what could be what could AI do to us and what could we do to them to like prevent like it happened right going out of control of us. Yeah also Buckman Center as well as MIT media lab have collaborated on the same like they have signed an ethics governance that this could be the steps we could follow to have the ethics on. So again there is a small help that AI has done on lip reading. So this is the online algorithm where the video goes on and so this is developed by Google where it will read the leap of the so it is even more accurate than what even we can do the one that the who is deaf so it will tell you the exact words which the like speaker does. And yeah there are small other three of two of the three I think news that I found out very very recently the one just a week or two before like Google already developed a small neural network architecture where it can store the entire short memory and long term memories also. So like using those they have some made a small robot which which have directly gone to the entire every corner of London with only his own memory. So every time like he started with a small it started with a small like zero memory and then travel through entire London and learn what exactly the map is. Yeah thank you any of the ideas that you have and what to share with people or any of the QS. Recap of the past present and future states of machine learning. Quick question states we do have to make a room. This one. Yeah so actually what happened is like there are most of the acquisition was done by two group one is Google and one is Facebook. There was one small letter from government that we could stop developing AI and they will not be issuing a lot of certificates or lots of license I would say to the new upcoming AIs. So what Google and Facebook offline did is they didn't disclose that they have already like signed a merger and acquisition with small companies and government has also increased the tax only to stop the. Going around in the public so this is so after the third one it we could have anticipated a very large peak also again but a lot of the news didn't came up. So if you see here the deals are the same a same number of deals 155 but the number of the money has decreased. So lot of lot of data was offline that's why it is written disclosed funding so then they didn't disclose lot of funding in the last quarter. So this is the second one actually and this is the third one for the smart city in what exactly I think not exactly AI not artificial intelligence application exactly. But I have seen lot of predictive like applications which are not exactly using AI till now. So it was still kind of like drawing a line around the data. Yeah, so just like small fitting line around the data but not very advanced artificial intelligence application. Any you know I think Japan might have but we are like out of news from Japan and China for like a lot of decades. We don't get lot of news from them. We might anticipate that we can get more projects from them. So is it already production? Yeah, last year. Thank you, thank you.