 Hello. This is quite new, bots and AI is quite new, and I guess all of you guys are excited. So there's going to be two sessions, as Santosh said. So mine is more overview kind. We'll just see what's the scenario, what's happening while it's more hands-on with Santos. So people think like AI and bots just started. It's actually not so. There's a lot of hype since recent days, since Mark Zuckerberg announced that they would support bots in a messenger. So yesterday, I had lunch with a friend who is also a starter founder. A couple of months back, he started a fintech company to do V2P lending, and yesterday we had lunch and before we had lunch, he made me sign a NDA agreement. So I said, what's this? I mean, I know him since last five years, I never signed the NDA before having lunch. So it's like, oh, I'm going to start something absolutely interesting, absolutely groundbreaking. I said, okay, let me sign the NDA, then we had lunch and he says, oh, I'm going to build a bot for fintech. So what he plans to do is he plans to do like a money lender bot, so people come and chat with it, and the bot decides whether to give money or not. So it's- How much can you see or just revealing that he- I don't know. So basically, I mean, I said that because there's a lot of hype these days around bots, people think bots can do anything, and bots can get your investment. So if you're doing a startup, just add bots to it and you will raise a million. So I think it's just a hype. I mean, part of it is hype, part of it is true. There's a lot of things happening. So I think like through my slides, we'll analyze how much is hype and how much is truth. None of us are experts. So if you guys know more, if you are doing interesting things, please chip in, it's a discussion forum. And it's like we discuss as we go. So the things we are going to cover is, we'll just see what's the chatbot, what's the history behind it, and why is this chatbot hype happening right now? Who are the major players, some of the use cases, and how is the future going to shape up like? So what is a chatbot? So a chatbot is just a piece of software that is designed to start conversations with you, to talk to you as if it were a human being. And mostly the use cases would be like, to answer queries, to get certain things done, people. So if you think like chatbots just started, you are wrong. So the first chatbots was developed actually 50 years back in, so that would be when? 1966, yeah, and it was developed in the MIT Artificial Intelligence Laboratory, and it was called ELISA, and this is how it looked like. So those days we didn't have windows and you didn't have graphical user interfaces, so you had to go to the DOS prompt and actually type all these things. And it was like text-based. And the way the logic worked it, it was very programmed. Like if someone asks you this, you answer this. And the main use case of ELISA was for entertainment. And these are the kind of the conversation people used to have with ELISA. People just used to like play a fool around with it. And so many, many years later, there was the next version of the chatbot, which was Alice. So apart from talking, it also had avatar, so you could actually see someone move around and smile at you when you talk. So when I was a kid, I used to actually talk with Alice, like, you know, and that time also there was a chatbot hype. So this is like early 2000s. And people used to put such bots on their websites. So when you open a website, someone will start coming and talking to you in voice. So that time there was a hype as well. But then after a couple of years, this hype died down. And no one uses, no one puts such avatars on websites these days. But then like, you know, this trend has been growing. As in, so artificial intelligence and virtual agents, like maybe 2005, there was again a hype of virtual agents. So that's when like technologies like Jopim, live person and all came on, where you're like, you know, if you go to a website, now either is you find a support button, talk to us button and you can click and talk to it. Some part of it is like automated. And this thing started happening in 2005 onwards. And the last thing is none of the chatbots have passed Turing's test. So what is Turing's test? So Turing's test is like, you know, so if you build the chatbot and you want to pass this test, what you have to do is so you have to create, so you have to create two instances. In one case, a person actually replies. And in the other case, your bot replies. So and you bring in a person who doesn't know whether it's a human being or a bot and he will not be able to like really recognize. So none of the chatbots developed till now actually have passed this test. That means it's always possible till now for a human being to identify this is a bot. So basically it will be interesting when things develop further and actually the bots pass this test. That means it's no longer possible for a human being to actually identify, oh, you are not a human being. So the way like the current chatbots that work for most companies is like when a person talks to the bot, this bot, say if you talk through Facebook. So you immediately pass your Facebook ID to the chatbot. So the chatbot now knows who you are. It can pull out whatever previous conversation it had with you, if you had talked to it before. It can also pull out data from, if you're talking to a business, that business might have some history about you. It might have some, you might have purchased things from the company before. So these things it can pull out. It can then match patterns. If there's some patterns the company has filled in, it can match the patterns. And then it can have some intelligence so that it can predict like what are you trying to do? And based on that it drops the response and sends back to you. So the advantages of chatbots are like, there's nothing to install. No longer you need to download an app which takes you one minute or so. It doesn't take up your phone's memory. The UI, once you start a conversation with the chatbot, it actually remains. So if you have Facebook Messenger, you start, it actually remains there. You can get back to them. They can get back to you at any time. It's more social because people are getting used to chatting. So this is social where you are actually talking to the company. It's no longer, it's a more, you have a stronger relationship with the company. It's leaner. You don't need to build heavy apps. You can just build simple and the interface is chats. And the main advantage is there's a lot of services. You can bundle into a single app. So if you're a Facebook Messenger, there's no limit to the number of businesses that can come on top to it. And you can just search for the company you want to talk to and you can just talk. So the chatbots, again, as I said, is not new. So if you do a search in Google for chatbots and you try to see the trend, you can see actually in 2005, the hype was even more than it is in 2016. So it's a sustained interest that people have in artificial intelligence and chatbots. So and it's going to get stronger over time. And to the reasons, why is this chatbot revolution happening right now? One is like app fatigue. Like maybe a decade back, people started building apps. Apps were new. Apps could be better than websites. You could get things done faster. But then as more and more people build apps, there's app fatigue. So people don't want to download apps anymore. People want to download few apps and get things done through them. So if you are planning to build an app, you really have to prove that a person actually really needs you again and again. That's when the person actually downloads the app. So if you are just building a MeToo app, the people don't download. App stores are getting crowded. So if you're building a new service, it's hard for people to find you. And you have to, so one is like the development effort for apps is too much. You have to spend a lot of time developing iPhone app. You have to develop an Android app. You have to develop Windows app. And then you have to keep updating it as these platforms develop. So with the messenger, with the messaging bots, you don't really need to worry about building different versions. It's purely text. It's simple. As the person updates his messenger, you can have new updates automatically happening. 2015 was actually a historical moment for messaging. The amount of conversations happening in messaging apps actually surpassed the amount of conversation happening in the top four social networking apps. So everyone is like, the amount of time people are actually spending on Facebook is much, much less than what people, the time people are spending on WhatsApp these days. So it's like, I was talking to another friend of mine yesterday and he was telling, oh, if you are not on my WhatsApp list, you are not my friend. So it's like Facebook, it's like you go add everyone up, but then people select who are the few that they actually actually want to talk to. So they are on WhatsApp. So that's the difference between, in social network, you want to keep it as wide as possible. And while you're messaging apps, you want to minimize to the real people you really care about. The other thing that's leading to this chatbot revolution is strengthening AI platforms. Like, so most of the big companies like IBM, Google, Facebook, they have been working on artificial intelligence since many, many years. Actually, if you see, if you ask me, which is the best bot today that I have interacted with, it's actually Google. So every day people go and ask questions to Google and it replies and it remembers what you actually talk to. It doesn't look like a bot, but actually it's a bot. As in, it doesn't look like you know, you are chatting with it, but then people actually keep searching, people search for one thing, then you search for the next thing, you search for the next thing. So it's actually artificial intelligence that goes behind it. And although it doesn't look like a bot, it's the best bot that's available for us right now. So the different technologies that they were working on are actually coming down from the laboratories to real world in the form of these intelligent startups, AI-based startups. And this is like the number of new prominent companies coming up in the AI space over the last couple of years. So you can see there's a strong growth in 2010 to 2015. In the artificial intelligence landscape, you can see these are the different kinds of companies that have come up over the years. So when it comes to bots, there can be two kinds of interaction. One is through text, one could be through voice, like Siri, Google Now, you can interact through voice. So part of it is the speech processing. Also, there are some companies who are working on avatars. So basically you don't talk to someone without a face, but actually there's a character who actually when you talk to it, when you can actually see the person actually smile at you and do things. There are companies working on face and gesture recognition. There are companies who are building platforms like NLP and machine learning for other companies to use them and build bots. There are companies who are building virtual agents and customer support bots. So basically if you're running a e-commerce business, you might need a customer support bot. This bot actually talks to all your customers about if they have complaints, it addresses those, and there are personal advisors, people are building bots which can actually advise you on stocks, which stocks to trade. If you give the example of customer service, it is known that a lot of people get frustrated. And there were high-profile cases where companies had lost a lot of PR points because of transition from human to both places. Yes, sir, what would you... Yeah, so just last week, there was this conference, Innovation Unbound, and there was the CEO of LivePerson. So LivePerson is actually one of the leaders in customer support, the chat-based. So what he said was, so definitely the bots are not good enough right now at this stage to really replace human beings. So a hybrid could be actually put in place where the bots actually face the human being first while the human beings actually monitor. They can take over the bots anytime. So that means like in most of the cases, like say 70% of the cases, the bot can actually successfully answer the query. If the bot thinks that it's going to face trouble, it's not going to answer, it doesn't have the answer, or the human being thinks that the bot is not doing a good job, they can jump in anytime. So right now at this stage, I mean, in future it might be possible that bots actually replace human beings, but at this stage it could be, if you want to build an intelligent solution, it could be a combination of both human beings as well as intelligent software. So where the intelligent software, intelligent bots actually try to answer the queries and if they cannot, if they think that the person is not getting satisfied, then they just pass a signal to the human being and the human agent actually takes over. I think that's the best strategy as of now. So people are expecting maybe it will be like 70, 30, as of now, so 70% of the queries can be answered by bots, while 30% actually comes back to the human beings. So it reduces staffs, increases productivity, and this number can go increase higher as this technologies mature a lot. So this is the AI landscape. So if we go to a bit narrow down to the chatbot ecosystem, these are some of the players in the chatbot ecosystem. So the right-hand side you find mostly the messaging apps. So that's where the consumers come in and try to reach you, while on the left-hand side you have the messaging apps. So that's where people come and talk, while on the right-hand side you have the companies who actually do the task. And in the middle you have the companies who are building the bots. So for example, you have Facebook Messenger on the left-hand side and you have like magic in the between. That means like a person comes to Facebook Messenger and chats with magic, can you book me a taxi? So then magic can actually link up with Uber and the Uber taxi actually comes for you. So you can see the three verticals. So one is the messaging, one is the input, where that's where the people come and talk. Like for example, you have Siri, you have Google Now. So you can come to Siri and ask, I need a taxi. So then the bot actually lies with the providers, like say TaskRabbit. So TaskRabbit has freelancers who can do things for you. And this is how it works. So when you're building a bot or you're building a platform, you can identify yourself with one of the three categories. The investments that are happening in the AI and bot space is also increasing. You can see like over the last two, three years, it has like almost doubled up. And you can see here, what is the investment actually going into? So a major part of the investment is actually going into the companies who are actually building the platforms. The infrastructure, which will be used by other companies to build their bots. And all the bigger companies are actually rushing to acquire the companies who have been working in the artificial intelligence space. For example, like Facebook acquired with.ai. Now with.ai is a part of the development framework that you can use to build Facebook bots. Similarly, every company like Google, even Yahoo has acquired a couple of companies in the AI space. Amazon, Intel, everyone has tried to grab some of the other company, which is in the space and doing well. So that they, if you see what Mark Zuckerberg said in the last F8 is like, in their next 10 year timeline, one of the key areas they're going to focus is on AI. So because like all these missing apps are becoming platforms and as businesses come, some kind of automation is necessary. And people want to spend time on these platforms to get things done. So that's why like AI and other things are necessary. That's why there's a rush for acquiring companies. So the major players who are actually providing the platforms where you can build bots are, I have just listed some of them. One is like Slack. So if you are a developer or you are a business, you might have used Slack to communicate with your team. Since people spend most of the time, I mean like say you use Slack, you always have Slack running on your desktop or on your phone. So people want to maximize the things that they want to get done through Slack. So for example, you want to do a deployment, you want to get it done through Slack quickly. If you want to get notified, if someone tries to hack your system, you try to have those notifications come on Slack. Similarly, Telegram has also opened its platform to bots. Actually, Telegram is leading in the bots space. I mean, Telegram has much, much less users than what WhatsApp has. It's mainly used in like Russia, Ukraine, and those countries. But like, so it tries to get an advantage over WhatsApp by taking a aggressive approach on bots. So if you go to Telegram, you can actually find all kinds of bots. I installed quite a few. You have like hotel discounts. So this bot actually gives you discounts if you can choose where you want to go. In Singapore, there's a craze about gambling and 3 or 4D. You know, you have like Lucky Bot, which can give you 4D numbers and you can go gambling on it. If you want to play simple games, you have the Tic-Tac-Toe and other games that you can image bot. There's a bot which can actually get you images. So, and the best part of it is like, now you have a bot which can actually create bots. So this bot is called Botfather. And you can just go, I'll actually show you. You just type new bot. So it already created a bot for me. Now it asks, what's the name? But like, you know, as the conversation goes on, it actually creates a bot for you. It'll ask you for a keywords and without actually doing any development, you will have a bot ready. What does the bot do? Sorry? What does the bot do? No, initially, you can just give keywords. So saying like if a person types this, answer this, but you can actually, so they have a platform open. So you can get a developer to build a stuff for you so that the bot does actually useful stuff for you. It just creates a basic bot for you so that it's ready for you to start. You can build more complicated things later on. So it's hosted? No, so the simple one is like, they have a directory with some of the keywords and all. So that part is hosted, but if you want to do like, you want to, you have a e-commerce business and you want to do more complicated things, you have to host it on your own. So there's assist bot. This is mainly in the US. You can see it can, you can do a lot of things with it. Like you can get a ride, you can get food delivery, you can reserve a restaurant, you can send flowers, you can visit places, you can book tickets, you can book hotels, you can send letters. There's this bot called dinosaurs. Every day it sends me some news article about dinosaurs. This bot sends me horoscopes. People who believe. This bot sends me maths puzzles every day so that I can solve and develop my mathematics. There's iClinic bot, which actually, so I complained to it that I had a headache last week and it found some doctor for me and gave me some advice and medicines. So you can visit this website called storebot.me. It's the telegrams app store for bots. There's a lot of bots in different categories. You have games. You have the bad bots as well, the last one. Yeah, there's this Andy bot, which actually every day it tries to teach you grammar. And the other thing is like telegram has just launched a $1 million developer challenge. So if you're building bots, you can apply for it. And I think maybe it ends in two months time. So you might be the lucky one to get a million dollars. I think the most interesting and the most impactful was Facebook's announcement about AI. So if you see in the 10 years timeline, I think AI is the focus for Facebook. So a bit about Facebook bots. So Facebook allows, Facebook gives you this messages buttons. These buttons you can put on any website. You can have this button on your Facebook page. So that what this means to the consumer is that they can talk to you on Facebook. That's how the conversation actually starts. Facebook also allows you to create this kind of codes. So you can put this code on a product. You can put it on, like say you are doing this event. You can put it on your door. People can scan this code with your messenger and talk to you. And also like if you have a bot, Facebook puts the bots category in messenger so that people can easily talk to the bots that they have recently talked to. Also Facebook does something called customer matching. So this is right now available only in the US. So that means if you have a person's phone number or email, the business can actually initiate a conversation with the consumer. So for the bots, mostly it's about the person initiates the conversation, not the bot. So Facebook doesn't allow you to go spamming everyone. But if you have the phone number of the person or the email of the person, you can actually be the first person. The bot can be the first person to start a conversation. But this is available only in the US and it costs 99 USD per year. So the messages that you can send in Facebook can be text, can be images as well. You can have structured messages, that means... So the current scenario is the NLP and the AI technologies are improving, they are good, but not good enough. So there's a lot of chances that the bot can actually go wrong. So the companies are trying to minimize the chances of failure of a bot. So by giving options. So that means so you want the person to actually choose only one of the three things. So that's where you give the structured messages. That means you are telling a person don't type anything, just click one of these. It's easier for the bot to understand actually what option did you choose? Sorry? Yeah, okay. Facebook also provides you the WIT.AI framework. So if you want to build an intelligent bot, you can use this. You don't have to develop your own NLP engine or anything, you can use this to create a bot. There are other bot platforms as well, like Chatfuel, Pandora Bots, which you can use to create your own bots without actually reinventing the wheel. So the current scenario is bots are getting popular right now. More and more people, businesses are thinking. So I just heard some a couple of marketing agencies talk last week. There was a round table. Do we have a bot strategy? Like earlier, like a couple of months back, couple of years back, people were saying, do we have a digital strategy? People were trying to move to digital. Now, do we have a bot strategy? So basically, bots are getting significance. People are thinking about them. Should we do something about it? And people want to come, but businesses have concerns as well. What happens when you build a bot is you build a bot on WhatsApp or Facebook and you expect your customer to actually come there. So that means you are actually giving your customer to Facebook. Like there was some discussion forum about building bots for fintech. And so then the concerns that these people, the finance companies were had is what if Facebook itself starts a finance business? What if Facebook itself becomes a bank? Then it takes all your customers away and goes with it. So people are still not sure. People are evaluating what's their bot strategy. So it's the right time for... So if you're a developer, you have some artificial intelligence knowledge. You want to build something. You want to build bots. Or you have a entirely new business model to do something. This is the time to actually start and get visibility. And bots can definitely simplify operations like filling forms, collecting data, getting opinions. There are many limitations of chatbots. One is like if you want a bot to be like really good, you want it to be close to passing Turing's test, which is like being close to not being identifiable whether it's a human or a bot. You really need significant computational complexity. The program behind it cannot be very simple. So the more complex it is, the more real like it is. The other complexity is like languages. For English, it's getting slightly easier. What if people come and speak in Chinese? We have a customer in Cambodia. And half of the people speak in English. And some of the people speak in the local queer language. And people mix both and speak and chat with you. So you don't know what is the person actually talking? Is it Chinese? Is it English? Is it Chinese English? Or what? So I mean the more the permutations and combinations come, the more complex it becomes for the bot to actually identify what the response should be. And people have hundreds of ways of asking the same thing. So I mean the chances of failure is quite high there if you haven't programmed the bot properly. And like you know, bots don't recognize humor and sarcasm. So they might take it for the actual. And so how about the future? I mean, this is like just my personal thoughts. People, so when Facebook announced and so every news article was writing, oh, apps are dead. You know, bots will replace apps. I personally don't think bots will ever replace apps. Apps will have their own, they have their own strength. Bots have their own strength. So I mean they will coexist. Bots will be used for the things that where a conversation actually matters. I mean if you have to think like you want to book a taxi or you want to consult a doctor or do some common things, doing it through a chat, you will have to talk five lines to actually get to the same thing that you can actually click a button and get done through app. So there are some advantages of bots. They will be used in that case. There are some advantages of apps. So both of them will coexist. People will start building, so right now most of the focus is going on building bots on Facebook Messenger, WhatsApp and other things, but there's also a potential of putting a bot into your app. So whatever your app you have, you can put a bot into it and it can actually guide people about how to use that. It can help people when people have problems. There's an interesting company in India called Help Chat. I don't know how many of you have heard of it. So since people are coming and talking to businesses on messengers, so this company Help Chat, so initially when they started, they didn't have any business sign up. So they started promoting themselves, so they built a messaging app and this messaging app's focus is answering business queries. So they will answer, so if you have a query about a particular telco, a particular TV channel or anything, they will answer it. So they created channels for every business that they could, so people came and started answering. So that time they didn't have the businesses signed up. So they put like around 100 to 100 people, these people started answering questions, human beings now know what. And as time passed like maybe in a half year or so, a lot of people came and started asking all the queries whatever they had in this app. So that is the time they started going to businesses saying, hey, do you want to come on board to this app or not? So now businesses started coming and being providing support staff behind this apps. The other thing, I mean, as I said previously, like I said that the best bot available right now is actually Google, which doesn't look like a bot. So in the future, I think like the best bots will actually be not visible. You don't know it's a bot, but it will be in the background doing stuff and communicating with you, sending you things, understanding you. I think those are the cases which I think will succeed and will grow to greater heights. Yep. So for my company, so I ran a startup called Noise Street. It's a five year old startup. We have been focusing on the media marketing advertising industry. We build products which are used by companies like SPH, MediaCorp, some of the big guys here. We operate in multiple countries. So recently we have been focusing on AI and bots as well, but ours is more product based. So we build solutions which a lot of people can use on a software as a service model. So we are building intelligent customer support. So this customer support is like a hybrid. So you have both human beings and bots. So if the bot tries to answer 70% of the questions and if it can't answer, it signals it to the support agent, the human being and the human being can take over. We are trying to build simple games over messaging apps. And also using that for marketing campaigns. So what happens is we, I mean over the past four, five years we have been doing a lot of marketing campaigns with retail brands. And so if there's a physical event going on campaign, a lot of businesses try to make people download apps or visit a website which is slow. So things can be a lot improved if you ask people, hey, add us on Facebook Messenger and say hello to us and you know the conversation starts and they can participate in the campaign. Bots also have a lot of strength, a lot of potential in actually educating kids, simple maths, science. People like kids especially love bots, they love to interact with them. So if things are presented in a very fun way, the chances of education, learning and people liking it is fairly high. So like learning for kids is another area we are focusing on. That's it. If you have to contact me, you can just drop me an email at helloandnoisery.com. Any questions for me? Or maybe we could start with Santos and later we'll. Yeah. And this chatbot AIs will be fused together with WhatsApp. Right now WhatsApp doesn't allow that. But very likely it will be opening up as well. So there's a couple of companies in India who try to fake, not fake, as in try to overcome this by they wrote some scripts which actually talks to you on WhatsApp, but WhatsApp has been very aggressively trying to block them. So that couple of companies who change their numbers every day and at the end of the day WhatsApp blocks them. So right now not possible. You said to start a conversation, you need to have their email. That's only available in the States. Do you know if it's gonna open for the rest of the countries? I don't know yet, but most likely. And that's even though the person opened a talk with me, I can't reply to them tomorrow, for example. No, so if the person actually starts a conversation with you, with the bot, the bot can reply back. At any time. At any time. Yeah, we. Yeah. So I can still do some notifications. Yep. Is it possible to, okay, so I have my own messenger account, a Facebook account. Is it possible to put a bot in front of it? Or do you have to, you know what I'm saying? Like if I have an account already, can I, is there APIs for me to make it into a bot? So you have a personal account? Yeah, my personal account. Personal, I think it needs to be a Facebook page. So people come and talk to Facebook. So you need to do the $99 thing? No, you don't need to pay. You need to pay if you want to initiate the conversation. But then if a person comes and talks to you, you can reply back. No problem. Okay. So that $99 is just for that. And it's to prevent people from spamming. So basically what happens is that it's not there. I mean the $99 will also come with you feeling reforms, identifying yourself as a company and whoever. So that it's not that easy for people to just come, spam and go. So it's a general bot, I don't think. Yeah. I heard that WeChat is pretty mature. Yeah. So Facebook actually is behind WeChat in terms of functionalities. In China, WeChat is free. And all these platforms are open. And anyone can develop all these things. Why is that? Line just opened up. Earlier it used to cost quite high to actually have a line official account. So right now you can create a free line account. But I don't think the developer platform is open to developers yet. You need to apply and... You need to apply. There's approval process in many ways. Yeah. But it never gets through as in... I mean they're trying to streamline things. I mean with all this competition happening, they're trying to speed that up but it happens slow. Any other questions? Okay.