IBM Watson - Artificial Intelligence: The Reality Behind the Hype

We are about to start. Thanks for coming. There are some people that we want to thank for tonight, especially our sponsors. I want to go through real quick. Mind the Bridge is hosting in the space here, as well with the support of Soma Center, which is the building and the co-working space where we are. To give you some ideas of why we are in the basement, what we are doing here, we have three stories building. The two floors that are upstairs are for startups. There are different startups from different countries, a lot of diversity upstairs. Very interesting, as usual, co-working space. This is the place where we used to throw events. Thanks for coming again. Mind the Bridge is hosting. We had a good partner, which was Vansi. We have a good partner in terms of events and marketing sponsorship. Then we have a good partner, which is our guys. They are doing the livestream tonight. They brought a very nice bicycle with all the tools that a livestreamer and a TV needs to stream a video or something. I want to go through real quick why you guys are here, what are we doing, who we are. You have seen so many returns around about Founderspace, Mind the Bridge, and other companies that are upstairs. We are building a niche of people that are working in AI and marketing. Who we are is me, I'm the founder of Aima. We are an A team, a serious A team. Somail is my co-founder. He's going to speak in a second. He's going to give you some details about what is Aima and what are the next steps. Then Jean Lombard is front of the air. She is our PR and communication manager. We are going to have some words from you later on. Pretty diverse background. I've been starting working back in time for Prezi, doing some community activities back in Italy with Uber as well. I'm not sure if Uber is the right logo to put here right now. But, yeah. I've been coaching in HALT, International Business School, and right now I'm working as Mind the Bridge. This is my daily job. Somail, I've been working in Apple, Motorola before, TiVo, and Apple. Two companies, he took two companies, IPO or Exit? Exit. Then Jean Lombard, very good experience in PR and communication, especially in AI company. What is Aima? I want to give my words right now to Somail. My co-founder is going to explain to you what is Aima and what you guys are doing here, why we are throwing this event about AI, and what is the reality behind the hype that all of us are seeing outside. Thank you. Thanks, Federico. Thank you all for coming. We really appreciate that. I know it's a long weekend and so hopefully everyone is awake. Federico and I have been talking about artificial intelligence marketing, some of the challenges, and we put together about six months ago just a kind of working group to see what the challenges are. And then we realized that there's a lot of opportunity to really bring a community together and really understand how we can build something that is for AI and marketing. And so we're going to talk about what it is, why we're here, and other than having a free food and beer. So here we go. Look. What we try to do is we try to build together marketeers and AI experts who are interested in really building a vibrant community to learn about AI. Some of the things that I didn't know a few months ago, I've been learning quite a bit. We wanted to figure out a way to share that knowledge so we all understand that. The other thing is that also we want to be ultimately becoming a voice of AI marketing. Together we can basically communicate to the world what it's all about. The most important thing we want to do is also to help some of you guys who are in large enterprises or in startups to really start building those partnerships and working together. And so hopefully that becomes a little bit of more networking sessions with the beers. In terms of why, we wanted to figure out different definitions. As we were talking about these things, we were seeing different definitions. We were seeing different vernaculars coming out from people about what AI marketing is. We have generally misconstrued deep learning to machine learning to just basic data mining. And we wanted to figure out a way that we can come together as a community that we can define certain types of vernacular that we can all talk about. And the resulting thing is that we want to have, you know, all of these things what's happening is that we've been seeing, even some of the companies I work with, where the AI marketing has been slow in adaptation. People are a bit fearful of not knowing what it is. Or not being able to figure out what's happening. And that's the difference between startup A and startup B. And so the goal of all of this is that as you guys come together and we all want to participate, we want to feedback from you, we want information from you, learning from you, so that we can tally all of this together and really create an industry wide, you know, something a little bit similar to what IAB does in marketing and AI. So in terms of, you know, as we were thinking about which areas we want to tackle, there's so many different areas. And we decided that there are six areas that are really interesting for us. And so the first one is media buying and selling. You know, we do this through Facebook ads, Google ads. And so as these companies are talking about AI, what does it really mean? How do you apply that? What are the different types of techniques they're using to make it beneficial? So that's the one vertical or one stream we're going to be talking about. The other one is going to be content and PR. A lot of these new tech companies are coming out talking about how we can automate certain PR, certain content writing that can be done. And so what are the things that are going on in the marketplace? We're going to be talking about that as another stream. The third one is something that's been happening quite a bit in PR market here. You'll probably know this. Fraud. So there's been a lot of fraud going on in different types of platforms. And there are different companies trying to figure out how to deal with it. Well, how can AI help that? How we can improve on that so there's less fraud coming in? The other one is analytics. So a lot of us create a lot of data and we have dashboards. The challenge that we see is that how do you take that data and how do you make something meaningful out of it? So how does the product manager kind of figure out saying, here's my data stream, here's what I'm getting, how can I build new products and features? From a performance perspective, what I can do to improve that. So what are the different AI techniques we can use when it comes to using analytics data of your own platform or of your own solution? The next one is pricing. When it comes to media buying in general, I've seen bids coming at 50 cents to certain companies of bidding $20. Well, that's not true. The price elasticity is so much. So how can AI really help that in doing that? Now, whether it's for advertising, it could be also for pricing a SaaS product. So how do you price a product? It could be different from different times. It could be different based on the consumer that's coming in more frequently. So how do you make it adaptable? And what do you can do with AI to make that happen? So there'll be other vertical we'll be talking about. And the last one is segmentation. This is something that all of us have talked quite a bit about it in terms of targeting. And finding out that one golden unicorn that an advertiser is looking for and basically bidding for it. So what kind of things we can do from an AI perspective to identify the right people. And all of this coming in comes with a premise that what do we do with privacy? Data is great to have, but it's also important that we need to be taking care of the consumer and their privacy in a mode that works. And so the company I work for, we take that very seriously. And we talk about it quite every day. And so what can we do as an industry to really help that? So there's no data leakage. And Instagram just had a data leak last week. And that creates another set of problems. So how do we manage that against those kinds of situations? So these are the kind of six streams we're going to talk about. So as we go along, we'll be sort of approaching each and every topic with those six verticals. In terms of activities, I'll pass it on to Federico and he'll explain more about that. Thank you. Awesome. Thank you so much. Um... Just want to make sure one thing, guys. You all have the Wi-Fi, right? So the Wi-Fi is over here. The Wi-Fi is on the other side. Make sure, you know, the password is CaptainOff and the Wi-Fi name is Founderspace. So as we go forward, our activities are... We are going to throw out a regular series of events. Expert panel, fireside chat, use cases and research studies. We are planning to do specific workshops on challenges and what is happening in the ecosystem. So in the way we are building a niche, our objective is to really dig down into the details with the people that are part of the communities. So most of you are coming from our meetup, which is AIMA, which is Artificial Intelligence Marketing. Some of you are coming from other different meetups. All the people that have been participating for this event will be added to our community. And if we are mistaken somehow and you are not going to get added to our community, please ask me or try to find our meetup because ultimately, to follow these next steps, this is what you guys need. You need to be subscribed to our meetup to follow us and see the next events that we are going to throw out. Akathons is another... Here in Silicon Valley, it's a form of... kind of doing workshop in a different way, usually hosted by the company. We are going to do workshop in AI and marketing, very specifically. And ultimately, what we want to build is a total leadership behind all this ecosystem, which is... What I want you to think about right now is there are a lot of people outside that are talking about AI and the application of AI in marketing. But rather than talking about AI and the application in marketing, what we want to talk about is AI marketing. So be very specific. Before, people were talking about automation in marketing. Now they are talking about marketing automation. It's a very well-recognized niche with standards and the right experts that are leading the conversation. So what are we going to do together? Together, we will learn about AI marketing. That sounds pretty clear so far. We are going to go through the latest news and articles. We are going to build a stream of articles and news that you guys are looking and that you can keep yourself informed. We are going to work on case studies. This ultimately is one of the best parts. We have been throwing out another event already with three case studies. They were very expert people, a company with 250 people in Silicon Valley growing up very fast. And this is ultimately very interesting to see how the ecosystem is developing. Sorry. Research studies. We are going to go through research studies very deep into AI and marketing. Expert Q&As. Discover new niches within the AI marketing environment. So what we are going to do ultimately is see all the different digital marketing niches, which is the verticals that we have been talking about and maybe discovering other verticals that are coming up. It's not about what we want to build. It's about what the startups outside and the companies outside are doing. If there's a startup that is building a new product, working on a specific niche, and implementing some new technologies, we want to talk about that. And we want to tackle that thing and make sure that we are aware of what they are working on and they are explaining to us in the case studies format. And then end zone workshop. We are probably throwing out some workshops in AI in the next month. So if you guys want to keep updated on our meetup, you can check it out. We have specific workshops for people that want to start doing AI by tomorrow and they have no kind of background in coding up to the people that are more experts and they want to start developing. And the guy that is going to coach is called Francesco Mosconi. He is a very skilled person. He has been writing all the courses for General Assembly. This is our calendar. So far we are in August. We are looking to work until October. We are in September, by the way. Work until October and try to fit our verticals for until the end of the year. So we are going to work on media buying and selling, content and PR, and analytics. And ultimately, as we are looking for the future, what we want to do is building a full leadership, as I said before, and make sure that we have the right people in our community that can explain it and that can lead the conversation. So we are going to do this. And how you can learn it and share in this process. We are going to have member-sponsored case studies. We are going to have research articles. Those research articles are very important because we are able to dig into details about the market and about the trends in the market. And we are going to share articles. That's what we have been doing already. And so you can find articles outside on Medium, on LinkedIn. And we are going to use audio and video meetups. That means that we are going to translate, as we are doing right now, the live stream. We are going to put it online. And this is going to be a resource for people that can double-check it for the next time. And I want to give my word to Jean right now that can give you a sense of where we are located in the social ecosystem and so how you can find us. You want to come here? Hi, everybody. So social media is going to be very important as we proliferate our ideas, our community, and grow our community. So please do log right in if you've got your phone or your device. Ding, ding, ding. I think you guys all understand what these logos mean. Please go ahead and activate yourselves and activate yourselves for the community. Feel free to tweet, particularly feel free to tweet tonight about what you're finding, what you're learning, and what you're using. And what you're hearing as Michael Ludden, IBM's head of product, enlightens us on the reality behind the hype. Thank you. All right, we are going to call our speaker tonight, Michael Ludden from IBM Watson. I like presenting from the barrel. That's why I came, actually. All right. Oh, it's full of beer? Is there a... No tap? Oh, good for that. I actually like to have a beer when I present, too. There you go. Thank you. All righty. So the first thing I'd like to say is please, if you want to contact me, take this information down, take a picture. I don't have any cards, and I haven't reordered them. And you know what? I don't think I'm going to. I'm postcard at this point, which is really, if you know anything about IBM, wild. But this is my card. And I'm going to give a talk, which is a bit of a twist on a talk I've given before, called AI IRL, Artificial Intelligence in Real Life. It's, you know, I guess another title could have been the reality behind the hype, but I didn't want to go that bombastic. Normally I give a talk these days called AI in VR, and I heard the person a couple of talks ago giving a talk about VR, and I'm like, shoot, that's why I have to give this one. So this one's fun. If you ever want to hear a really fun talk, AI in VR is really interesting. A lot of what my team does is apply AI to VR, so we have an initiative called AR VR Labs. Yeah, take some pictures of that. But so I'm the director of product for a team that I founded called Watson Developer Labs. We build products for developers that solve problems they have in a use case-based fashion using IBM's tech stack, and we produce them on GitHub. They're open source. Anybody can take and use them for free. And the idea and the strategy behind that is to address new and even create new developer segments that we can then address and serve. So this talk, we're going to talk about AI terminology and just get to kind of ground truth because there's a lot of hand-waviness around the term and all of the terms actually these days. It's becoming a bit of a cliché in the same way that talking about cloud or big data or synergy might be. So we're going to try to define it at least. These are my working definitions. You don't have to adopt them, but I recommend it. Hopefully I didn't break my laptop. And then we're going to talk about different approaches companies are taking to AI. Companies like ours with cloud-based approach, locally trained machine learning algorithms using open source libraries like TensorFlow, etc. Talk about the benefits of each, the drawbacks of each. I'll give you a fancy chart. It's not fancy. And then I'll give you some advice. Yeah, so I didn't change that yet, but the previous time I did this talk was for startups, if you're thinking about adopting artificial intelligence today at your company. So first things first, what is artificial intelligence? What's someone's definition here of it before I go into these terms? Anyone, a brave person. By the way, there's a lot of audience participation, so I'm going to be doing this a lot to get used to it. Anybody brave enough to hazard a guess? Yes. Using a machine to do tasks that a human does. A for effort. A for effort. I think that could be very broadly, potentially. Yes? Yeah. Okay, so self-learning. Yeah, yeah, these are all sort of, what's that? What did you say? Yeah, patterns, knowledge graph, et cetera. So the definitions are all over the place, which is why this slide is here. We're going to get to ground truth. And you know what, none of you are wrong. All of you are right in some way, but what's a helpful way to speak about artificial intelligence? My view is that when we speak about artificial intelligence, we're actually talking about machine learning algorithms today. What's an example of applied AI that someone here has seen? And they're like, oh, that's cool. Yeah, go ahead. Netflix recommendations. Netflix recommendations. So it builds a profile based on the movies you like, and it matches you with other movies that it thinks you will like, and that's powered by a machine learning algorithm. That is exactly correct. So what is a machine learning algorithm? I'm not going to ask you. I'm just going to say it. A machine learning algorithm is basically an algorithm that can achieve a result in an unsupervised fashion, but it has to be told what to achieve. So an example of that would be in the 90s, Deep Blue. Anybody know Deep Blue? IBM's chess-playing machine learning algorithm that defeated the world champion at the time, Kasparov. It was created and designed to master chess, and it did. And that's the one thing that it did. And the same thing with AlphaGo recently. Anybody familiar with AlphaGo, Google's machine learning algorithm that conquered the game, the ancient Asian game of Go, where there are too many pieces or too many moves at any given time in the game for a single person or even machine to compute the optimal outcome? There are too many optimal outcomes, too many moving pieces. And so it was thought that we were several years off from a machine actually being able to beat a human in this game because it was fundamentally about strategy and intuition, things that we think, that we hold dear and we think are uniquely human. Well, not so much. So it beat the world champion, and it is now the reigning AlphaGo champion. And so in both of these cases, though, we're talking about a term that I would like to resurrect called narrow AI, which is to say a machine learning algorithm. And what do I mean by that? I mean that if you were to ask AlphaGo to do your laundry, it would have no idea what to do, right? Because it does one thing well. It's a narrow AI. It was told and it was instructed, you need to learn how to beat these humans. The objective is to beat them. And if you can achieve that, then your life is a success. It has no agency. It has no ability to transfer skills to other domains like you would not ask for dating advice, et cetera. And you laugh, but I mean, that's what we're talking about, right? It's not conscious. It's actually not even close to sentience. How this sort of works is it learns in an unsupervised fashion. There are various ways for training machine learning algorithms. Deep learning is one of those ways. Unsupervised learning, incentivized learning, but basically you can say via various different methods to an algorithm, I want you to achieve X result. And I don't care how you do it, or I do, depending on how you're going to teach it what to do. And then it just goes and it iterates and eventually finds a solution. So in the case of AlphaGo, what's really fascinating to me is in every case that it won, it was losing up until the last, around the last move, and then it beat its opponent by the minimum possible score. And the reason was, and this is fascinating to me, it says something about human psychology. AI frequently says a lot about us that's interesting and sometimes depressing. But basically the goal was to beat the other player, right? So what was the best way to beat the other player? Well, some theories say that it was probably to lull the opponent into thinking that they had the advantage so that they would back off. Like, oh, I got this, this stupid machine, whatever. Well, it wasn't playing for its ego because it doesn't have an ego. It was just playing to win. And then all it had to do was beat the other opponent by one point. And the easiest way was likely to bait the opponent into thinking that it had the upper hand and then to make a mistake and pounce. And that worked reliably well. So that's narrow AI. Now what's general AI? Imagine if you took out your iPhone right now and you're like, Siri, set a reminder for tomorrow. I've got a meeting at 8 a.m. with Brian or whoever. And Siri was like, fuck you. I'm napping. I told you never to talk to me at this time of night. I'm tired. It's been a long day. That's general AI. Does that make sense? Consciousness, agency, the ability to do different things, the ability to say no, right? That's general AI. And that is tied to the concept of the singularity. Anybody know about the singularity? What's the singularity? Consciousness with computers. Basically. Consciousness with computers. Yeah, that's more or less right. So basically it's like an inflection point at which artificial intelligences will achieve sentience to a certain degree. It's a very nebulous term, but basically, the thing we're all worried about, robots taking over the world, etc., that is that point, although it doesn't have to take the form of such a negative view of the future. We'll get into that in a little bit. So generally AI is what is going to occur when the singularity happens. And the singularity is either going to happen tomorrow or in 50 years or somewhere in between. We really don't know. It seems like it's close. But then again, so did chatbots 20 years ago, right? So you never know. Things could take a long time. Things could stall now. Or there could be a stealth breakthrough that's already happened and is operating in some compute cluster in Aruba. Who knows? Who knows, right? But when they start to be able to act and behave like humans and learn different skills and have their own artificial intelligence, and have their own opinions about things and form societies and interact with us and have their own language, etc., that is not taught and specifically directed by humans, that's generally AI, which is going to happen around the time of the singularity. Cognitive. What did you say? Sorry. No, I didn't say that. What did I say? Anybody want to repeat? and desire to do things. And when they're able to kind of achieve a level of sentience that... What's that? Desire. What is it? What is desire? This is getting real philosophical. Let's talk after. Let's talk after. I'm just going to quickly move on and we can, you know, hold your questions. Well, not hold your questions. Ask questions. But I'll talk to you after about that stuff. It's interesting. But cognitive is an umbrella term that companies like IBM and Microsoft and HP basically trotted out there to try to say, see, we're thought leaders. I'm not a huge fan. I'm probably the only IBM person who will tell you this, but it's an umbrella term. To me, it's completely interchangeable with AI or machine learning. Just like a cognitive system, a cognitive something. Oh, I'm building a cognitive application. Oh, what's cognitive about it? Well, it uses a machine learning algorithm to search your photos and show you pictures of your dogs. It's an example of how you might use it in a sentence, but I don't recommend you use it. Machine learning as a service. Just throwing this out there. Again, I don't recommend you use this, but if you see that acronym, it's just a past offering that features machine learning algorithms. Machine learning algorithms as a service. So, okay, current approaches to artificial intelligence. What we do, what Google does, what Microsoft does, what other companies do is a cloud-based approach primarily, although there are, we also dabble in local algorithm as does Google and others. Basically, the idea is the Watson Developer Cloud, which is our suite of services, are kind of black box, but they're accessible to developers as APIs. They're just instantiated as REST APIs, and you can make a call and get, for example, information back on the bit of text that you just submitted to the service. What did the person mean? What were their intents? It'll tag it with metadata and send you back a JSON file with all that good insight. Same goes for visual recognition, other services that we have. But the main limitation on this is that you need an Internet connection, and you're always beholden to another company to maintain and upgrade the service. It may not exactly fit your needs, but you're going to get the most advanced functionality out of machine learning algorithms available today by using a cloud-based approach, because companies like us have sunk billions of dollars into research and are always maintaining and upgrading the services and their accuracy, et cetera. Local algorithm. So TensorFlow is the most famous example, but it's important to note that at its core, it's just a set of open-source libraries that allow you to more efficiently train machine learning algorithms yourself. The time to value on this is extremely long, and I'm not poo-pooing it because we have our own, I think it's called Watson Machine Learning that we just rolled out in other efforts to support this, and I actually think this is a perfect fit for people under certain conditions, companies under certain conditions, and I'm going to go into that on the next slide. But importantly, you get to own your own algorithm that you train. It's custom-built for your purpose. Maybe you only need it to understand the word poo, and that's all you need is speech, you know, to text the algorithm to know. That's entirely possible. I heard of an example where NASA's building their own onboard machine learning algorithm for speech recognition for space missions on the shuttle using TensorFlow, and that's necessary because, of course, they'll have no Internet connection, right? So you can't use a cloud-based service. There's no Internet in space yet. Open-source is my favorite. I hope it becomes the sweet spot. I'm worried that it won't. There's a lot of small- to medium-sized startups and companies that are doing AI and algorithmia that are essentially open-source machine learning algorithm marketplaces where you can take an off-the-shelf, you know, computer vision API that recognizes puppies and put it in your pocket and put it in your app. They try to sell you a services layer around that and say, yeah, but you could get so much more if you work with us. And I like that because it leverages open-source and the community efforts around that, and it also gives a company that does this a first-mover advantage, right? If your competitors are going to use your solution, you're always going to be a little ahead of the game. But that said, the most immediate money, the most immediate money is right here, and that's why a lot of the big companies have jumped in the cloud-based approach. And by the way, GPU accelerated training, CUDA and the like, just exists. NVIDIA put out CUDA. It allows you to train using NVIDIA GPUs faster for TensorFlow Cafe and other open-source libraries that help you train your own machine learning algorithms than you would otherwise be able to. Because training machine learning algorithms, if you didn't know, has a lot of floating points and uses a lot of computational data that GPUs specialize in. Or, yeah. So, benefits of each. Cloud-based. Easy to implement. Quick prototyping. Free, usually. We have a free tier for all our services, so you could go and see what it would look like in your application if you supported a really advanced speech-to-text whatever, or whatever you want to do. Low ongoing maintenance cost. Obviously, it's nil, because you're not maintaining the algorithms themselves. You're just maintaining your application and making a call. If the functionality changes, of course, you have to react to that. And you get advanced functionality. You get the best, most accurate speech recognition, the best, most accurate, you know, visual recognition, et cetera. Data analysis, whatever, you know, sentiment analysis, whatever you want. You get all of that right out of the box, and you don't have to build and maintain it yourself for years. For local algorithm, you have your own ownership of it. You're not beholden to a big company. It's a custom fit, like I said. You can build it to exactly your need and nothing more. And that will save cost over time, potentially. And offline capable. So, if you're running things on embedded Raspberry Pis around the world, you don't need to check in home. If you don't have an Internet connection, you can actually have a computer vision API running on an IP security camera that's never connected to the Internet and still do facial recognition, that sort of thing. Open source is kind of the best of both worlds. The worst, the drawbacks of each, are cloud-based. You're beholden to a company like IBM, Google, whatever, Microsoft. Internet connection is required to make an API call, obviously, so you have to have a device that's capable of checking in when it needs to. Most devices are these days, but obviously, it's certainly in brick countries, developing nations, maybe not so much. So, service features may or may not fit your need. Services can be sunset if there's not enough people using it. Maybe it's key to your business, but maybe it's not key to anybody else's. It could go away, and then what are you going to do, right? You have to choose a competitor, or you have to have already built your own middleware, and think about that so you can easily swap us out with somebody else. Or maybe you'll have to build the machine learning algorithm yourself. So, you've got to think about these decisions at the very start. Local algorithm. So, the downsides of this are lengthy time and personnel commitment. I say this because it starts with one developer. You're like, oh, cool, look at this. But as you add functionality, and you're also thinking about, oh, we're also, by the way, competing in the marketplace of ideas where other people maybe are using cloud-based approaches, or maybe they're bigger companies that have been building their own algorithm forever, so it's an arms race and you're already behind. So, that can quickly balloon in terms of maintenance costs, especially as your customers get used to the goodness that your algorithm's providing. It can quickly get out of control, and so you really need to take that seriously when you're thinking about it at the very top. I actually advocate a bit of a self-serving mechanism wherein you can use the free tier here to see what it might be like while you're seeing what it would take to build your own machine learning algorithm, and if your company can support it and maintain it long term. You'll need a team of data scientists. You just will. And you'll need more and more people to maintain it as the functionality gets more and more complex. And the functionality you're going to get, by the way, is much, much more basic. So, if you're competing against Google Photos, and you can search for photos of my fiance from two years ago and get every single one, you're not going to be able to build that functionality in your own application overnight. That'll take years to do that remotely. It takes years to do that as reliably as Google Photos does it today. So maybe you just want to use their computer vision algorithm, or ours. Preferably ours, if any of you understand. And the downside is open source is also the worst of both worlds. And here is my fancy chart. Time to value is shorter with a cloud-based solution and longer with a local algorithm. Isn't this a terrible chart? So, oh, this slide's not supposed to be here. Advice? Okay. This is my last slide, and then I'll take a bunch of Q&A. Advice for how to think about adopting AI solutions. So, be specific about your use case at the start. What do you really want this thing to do? Do technical vetting to understand what would it take to actually build this. What is AI currently capable of? What can we build ourselves? How many people would need it, et cetera? AI is not a panacea. It's not magic. It's not that. We're not there. We're at narrow AI. We're not at AI that can tell you to go away. It's napping, right? That would be magic. Any technology sufficiently advanced, is indistinguishable from magic. Anybody know who that's from? You can look that up. Anyway, you guys should know that. My goodness. But AI is not there yet, and it is not going to solve all your problems. You can't just throw AI at something. I hear this every day, and frankly, we're a victim of our own marketing. The way we market Watson is way too high level. So, people think, oh, my gosh, I saw Princess Leia talking to robots. How can I make my robot do that? And we're like, well, hang on a sec. Let me walk you back. There are problems for yourself if you don't exhaustively research this before making a critical business decision as to what you're going to support. Also though, you're not married to your choice. Like I said, I recommend you start with one of the cloud-based free tiers, kick the tires, prototype, and then you can decide if you want to build your own locally trained algorithm, get one from an open source repository or marketplace like Algorithmia or H2O.ai, or opt for a cloud-based solution. And then do the math on the cost of maintenance. You have to have a roadmap, get your product manager to scope out what features you want, V1, V2, V3, V4, how many people is it going to take to support that, do we need a team of data scientists or just one, and for how long, because eventually you'll need a team. It doesn't get easier, it gets more complex. Yeah, people, money, time, it's not an equation, but I'm sure I could put that in equation form for future presentations. And then does this make sense long-term from a P&L perspective versus your competition? Like I said, if you're competing against other people that take advantage of or maintain their own cloud-based algorithms, you're in trouble unless you opt for one. Or you can exit the armed race and have your own differentiation. So that's it. Again, if you didn't get a picture at the top, feel free to take a picture of this and then I'll take some questions. Oh, thank you. It's hot in here. Right? No? Anybody have a question? Yeah, sorry. I know you've got questions. Yeah. Marketing automation is number one in marketing, obviously. But deeper than that, what have been some of the, I mean, In marketing? As in marketing automation, what functionality or capabilities have you seen take place as a result of ? Search, intelligent search, being able to do natural language queries for specific things for 40 other people being searched, that kind of thing. Being able to more easily segment. You know, your addressable market, your user base, et cetera. Detect, anomaly detection, data, that kind of thing. But honestly, it's much more basic than you might think. Like I said, it's not a fantasy yet. And you should be happy with that. You worked in marketing. It's still a good job. I mean, if it really was, it's not. Our whole marketing would be awkward. We would always know who to target and which one, right? But we're not. We're not close to that. I say that, but it could happen tomorrow, or some of you have a breakthrough. But really, there's sort of incremental things that make marketeers' jobs easier as they try to get to the value of the sense, more or less. And that may not be top three. I mean, honestly, marketing is something that I think relies a little too much on the arrival of AI. You still have to use common sense. And I think a lot of times there's a lot of weirdness going on, especially, particularly in the advertising space. divorce from reality based on what a machine is doing. Yes? Hi. I'm . What are the coolest applications you are seeing emerge for AI? I've been working in financial investing, learning. Your stocks are the best. And that'd be great. And . I've been finding data that's the most actionable. What's your favorite product? So you mentioned a good question. The question was, what are the most interesting applications you've seen using AI? Right? Emerging. Emerging? New. New. OK. OK. Well, I'm biased. Because we built some of this stuff. But I don't know. So Star Trek Bridge Crew has a feature powered by a toolkit that my team built called the BRS Group Sandbox that lets you talk to your team. And your team members that aren't actually humans on the bridge of the Star Trek enterprise in BRS play the game and have them take action based on what you say in natural language. You can use a wake word. You can use variations in phrases. And that's based on two Watson services with some open source code. And it's all in GitHub. It's freely available. You can use it. I think that's really cool. I'd like to, in fact, I think that interactive speech interfaces are about to get a whole lot more useful as we come up with that last mile. It's not just that I want Alexa to turn on the lights. I want Alexa to get the toast and the coffee. And post it. And put the jam on it for me. And give it to me. That's sort of the last mile that we're talking about. It's an analog world, right? And we're getting closer to actually useful voice interaction that isn't frustrating. It doesn't make you talk like a robot. And that kind of understands context. That's a big important thing. And it hooks into, by the way, some third party services. And we're at sort of the forefront of that, especially as it pertains to VR. Because that's essentially a fourth dimension of immersion. If you could speak to a world and have the world react. Whether that's a person, a system, whatever. And that's a whole other world. But another thing, you mentioned FinTech. And I will say this. We should all in this room be aware that these big firms are already leveraging artificial intelligence probably better than anybody else. To not your benefit, but to their benefit. So we're talking about high speed trading, understanding in real time where to invest and where not to. How to put together indexes. All sorts of interesting financial products that didn't exist before that artificial intelligence has come patterns around and said, oh, this could be interesting. And so we're talking about the market in ways that have never happened before. You're not really seeing that benefit unless it's a really good company that's doing well with your money and not sucking a bunch of fees out of you. But the stuff that you might see from FinTech that could be interesting is going to be the last thing to focus on. Like instant transactions and that sort of thing. But it is happening. And it's being applied very, very deeply in some of these firms. And I think that leads to a little more regulation, to be honest. It's strange right now. By the way, it's no longer a good idea to invest in individual stocks, period, at all, at any rate. You're just going to lose relative to other people that buy indexes and whatnot. And that's partially a result of artificial intelligence. I'll have to think about other examples. I'm sure something will come up. But I can't think of anything else. What about the neural? Neural. Like, say, for people who have handicapped LHIN or something like that. A little about the handicapped lens. You know, I think of VR and AI applied in VR. There's some great stuff going on with stroke victims putting on a few VR covering motion in arms where they've never had any, because there's something about VR that tricks the synapses in the brain to reconnect it, and that's awesome. I'm not sure if they're making use of the machine when they have a machine pattern. It's simply the act of being in VR and seeing an arm move when you think it. It's not when you take the machine. But that definitely is involved. But in terms of neurons, yeah, I mean, I guess computer vision, there is a lot of stuff that I've seen around augmenting reality for blind people where they tell you what they're looking at and speak it. It's a bit funky, a lot funky, but for many people it represents brain. Where they can never understand the surroundings, and now they have a personal key that whatever they look at goes to the blue pillar. That's all I'll say to that. Anything else? One more question. People are leaving, so yeah, go ahead. You were saying that the cloud-based solution is a much better solution rather than open source and user-developed, right? Do I think a cloud-based solution is a much better solution than open source? Yeah. The slide that you were showing, the time and value. Yeah. So cloud-based. A faster time to value the cloud-based services, but you can get some pretty advanced functionality out of the box with open source. The difference is that if you're going to fork it, then you're going to own it and maintain it. And there's some governance involved in that, right? Whereas with an API call, if the functionality is improved and it's more accurate, that's invisible. If you do the call to the same API, it just gets better. So that's kind of what I'm getting at. It's not better or worse, but I mean, I always say open source is the way to go if you can do it for any number of different reasons. Ownership, community, not being able to live in one company that's been sunset to service, the ability to fork if someone stops maintaining it. Okay, thank you guys very much. Have a good night. Thank you.