and 45 minutes. unenthk you for that usiastic welcome back. [laughter] i appreciate that. it's monday, so we are only going downhill from here. our next guest is very exciting to have him join us on the stage. we talked a little bit about ai. elon is not offend, yuri is. -- elon is not a fan. please welcome dr. john kelly from ibm and our moderator.

♪ >> we are going to do a live demo. what are we about to see, and why do we care? >> what you are about to see something that no one else has ever seen, other than those deep in my research labs. many of you remember watson, the artificial intelligence machine on the game of "jeopardy." at the time he was an open domain question and answer system. fast-forward to today, what you're going to see is watson has adjusted all of wikipedia, and not answers questions, but reasons over that and conform opinions on any subject. >> so the tech guys can make it turn on for us. the recount. i believe it has in introduction

going on first. watson? >> is he going to do with? -- to do it? is it going to happen? >> watson? please start. >> hello and welcome to the imb research demonstration. today we will research some of her keep abilities and generating argument for or against specific topics. to proceed, please select a topic. >> okay, let's go ahead and do on speech. >> skinning for one million wikipedia article. returning 10 most relevant articles. scanned all sentences. ready to deliver. thank you for the opportunity to share my view about this interesting topic. it is always good to start with some background.

the reliability of wikipedia compared to other encyclopedias and more specialized sources has been assessed in many ways, including statistically through comparative review, analysis, and historical patterns, and strengths and weaknesses inherent in the editing process unique to wikipedia. my opinion is that wikipedia is reliable. i would like to say a few simple reasons for that. contributors provide identification. another reason is that wikipedia provides better coverage and longer articles in general. these claims are supported by the scientific literature. for example, an early study conducted by ibm researchers in 2003, two years following wikipedia's establishment, found vandalism is usually prepared -- repaired so quickly that most users will never see it. wasoncluded that wikipedia surprisingly effective at healing. i hope i have convinced you to

support my view, wikipedia is reliable. thank you for listening. >> that is going to make high school kids everywhere so happy. >> so that is good. i am curious, when you in just that much data -- when you ingest that much data, had you formulate questions and answers? do you train it, or does it automatically understand how to pull up different topics from a data set? >> all you do is give it a topic. what you just saw, watson was not trained in that domain. we gave him the corpus of information and wikipedia. we gave him a topic and said, for many opinion on this topic. it went through an organized all the information. it understood the context of all of the passengers in wikipedia, took apart all the syntax and reassembled it and machine generated the language. >> how does it understand a

concept like honesty or accuracy? how does it understand this concept so that you can actually pick how to argue for some position? >> it looks for multiple scenarios and reinforcing information. for instance, i was surprised to hear a reference to imb research working in that area. found some was it passage in wikipedia, then it went off and validated it was legitimate scientist from ibm research before it would say it. it's constantly looking for verification of what it's about to say. >> is that piece of technology, is that currently something people can use in the market? is it more of a forward-looking thing that you have not released yet? >> its forward-looking which will be available as a service. let me remind you all that i see time of it jeopardy, watson was a large computer system in a room. since that time, we have taken watson apart, if you will, lifted it up to the cloud.

it's available as a set of services and apis. everyone in this room cannot go out and start to compose their own mini-watsons, if you will, in the application space. >> is it a six-month question, three-year question? how far a week away from having that out? >> this is probably 12-18 months from being in the market. that is a trivial example of the beauty of. remember that in the area of health care, watson has ingested almost all of the world's medical domain. if you give watson a disease topic, it will go into all andished pubmed journals form opinions on diagnoses or drug discoveries. >> i was doing research today and found a funny story. nikita khrushchev actually came to an ibm facility in san jose. this was back in the 1950's. they rigged up a computer to ask

questions. have you solved it? is it done now? or are you still making progress? >> that is a great question. prior to watson, everyone in the artificial intelligence community had tried to solve open to me question and answer by writing a month of rules. rules based learning, or organizing research. that had very limited success. it wasn't until we took a full open to me statistical approach -- open domain statistical approach that watson was able to answer questions in open domain. >> on that point, i have heard more about ai and computing in the last six months than the last six years. it seems that there is a moment going on in this space. why are we seeing innovation and growth so quick now? what has changed in our approach? >> we are at the proverbial

perfect storm. much of the world's information is now digitized, including natural language and areas like health care, law, etc.. computers can now access it. we now have the computing power to do something like you just saw. we did not have met a few years ago. >> just the raw capability. >> yeah, and what has really changed in the area of artificial intelligence is areas like machine learning, statistical analysis are now advanced to the point where we can do things like you saw there. that was not the case a few years ago. >> machine learning is one of the things watson can use. it's one of its building blocks. >> that is right. think of watson as a whole series of statistical learning engines. it's not one thing. and it's constantly computing and trying to learn over information. we do not program watson. we give watson data were better

data -- or better data, but we do not reprogram it. we give it better algorithms, but we don't reprogram how it operates. >> you give it a stronger brain over time, but you don't tell it what to think. >> no, it learns over the information. >> i brought some with us today. when you got to pitch developers watson in new york, what is your core pitch? >> the interesting thing is that it's such a powerful technology, i don't have to give a sales pitch. they immediately go into, how can i use this? can you give me an api to do a certain set of services? i was up at mit and they are blown away by it. they have been working at ai and machine learning for decades. they are just using the services that are coming out of this. we have opened up the platform.

there are hundreds of start up companies. ibm informed a $100 million fund to bring additional players onto the platform. >> so if you want a check, you should grab you offstage. >> middle of a progress in watson over the last 18 months. been impressive, has it been slow, or what you have expected? >> it has been exponential. we're running as fast as we can to keep up with demand. not just for developers, but for investors. even large enterprises are using it to transform areas of their business. old respectedry company. the history, lots of work. it's not snapped at or facebook. -- snapchat or facebook. do you think you have the share you need in silicon valley to attract other developers to your

company? or are you still fighting off that stodgy pocket protector kind of image? no offense. >> we do not have the mind share here. it's growing exponentially. but if you listen carefully later this week, you will hear things from us that will reinforce our commitment to the developer community. i would also remind you that when you think about the arrows of computing, the first was tabulating machines. -- the eras of computing. that was in the early 1940's. then we had programmable systems. every computer today, every device today is programmable. watson is not a programmable system. it's a learning system. it's the first of a third era of computing. i will remind you that ibm played an important role in the first and second eras of computing. we will play a key role in the third era of computing.

>> this is a good segue. who is her main come edition -- who is your main cover edition? >are there any big players tryig to encroach on your territory? there are a lot of great people working in the area of ai and cognitive computing. most are developing single algorithms for solutions to solve some problem, to improve search. >> a very narrow solution. >> they are narrow solutions. we are the only company that has developed a complete platform, open api, that ability -- the ability to innovate on top of this. if i had to make a prediction, for every innovation coming out of ibm research and development labs, there will be 100 or thousands of pieces of innovation that will occur on top of it. that is what we want. yes, we are in old company. for watson on a

regular basis? >> we do a value share platform. >> aws charges by usage. why did you pick that model? >> the first decision we made, we still could sell lots of watson boxes. but we decided we wanted to open a platform up and make it as a service, knowing where the club is going. -- were the cloud was going. that was a strategic decision for us. then it came to, do you charge by the click or not? because this is a learning system and is becoming smarter, it's important that we share the value with our partners and customers. >> have you had any pushback from developers? we want to pay on more of a click basis? >> no, it is a shared risk. for developers, you come on for

free, develop your business. as you are you revenue, then we sure that. but we want an easy on-ramp. we want to get going. we think is very fair. >> is watson going to be a key revenue driver for imb going forward? >> we have a set of strategic initiatives in the company. watson is part of our analytics business, which is part of a $17 billion double-digit high-growth business. watson is the fastest-growing component of our analytics business. >> now we can talk about the future. >> thank god. >> i'm going to presume that you don't think that ai is going to end mankind as we know it. there has been some talk about what this looks like, and should we be worried. do we build defense against this? this seems relatively benign and more industrial than progressive. >> when any new technology comes along, people get scared.

technology can do great things were not so great things. it depends on how we use and control it. whether that was the steam engine, automobiles, x-rays, you name it. this is a brand-new technology. it's going to change the world. there is no question in my mind. we think about what this technology can do, fundamentally if you look at decisions that human beings make, we make decisions with a bias. we make decisions on incomplete information. i believe that nearly or perhaps every decision that would make of any importance in the future will be made with a watson by our side. >> does not letting the computer making our own choices, it's giving options that we can select from. it's more of a addition. >> that is right. didsimple demonstration we around wikipedia. think about watson where it has ingested all the world's information on drug therapies for cancer.

i don't know about you, but if i had cancer, government, and the cancer board was meeting to decide what chemotherapy they were going to give me, i would want watson to have gone through it. to be sitting at that cancer board to augment the decision. say hey, guess what, you may want to rethink that based on this piece of information. or, here's a new con to the procedure. >> as watson gets smarter will they be more leaders in our lives? will we handle off stuff to watson? -- hand off stuff to watson? >> there will always be moral decisions, labor death, where it will be a joint decision. we as humans will always have the control of position. as time goes on, more and more decisions will allow the system to make. think about the internet of things when we need constant

responses in milliseconds. making that in real-time versus waiting for me to digest something in seconds or minutes. >> in our daily lives, in 5-10 years, does watson power experiences we already use? or is it a standalone product that i would talk to? >> it's going to be a combination. it will be literally everywhere. it will be powering things that you just take routine in your daily life. if you are buying something, watson will be behind that, enhancing your experience. it will also be visible in many of your applications. you are "if, or you want -- you are being creative, and you just don't have the time. it will be both visible and highly invisible, both ubiquitous in every thing we do. >> but to be fair, you want watson to be a brand name and technology.

will i have a watson sticker on my phone? >> watson is already the premier brand in ai. not something we set out to do, but through the demonstrations in jeopardy and whatnot, they have already had brought -- broad recognition. as we move forward, we will decide whether we want to do "powered by watson." we opened up a new division of ibm around watson. >> health care vertical. >> stand up a new divisions around education, the internet of things. as watson becomes capable of operating in those domains, we will bring him into those areas. >> before we go, humor me with this last question. was there a moment where you and wow,team were like "oh that actually works." how did i come to be? -- did that come to be? >> on the journey to jeopardy

back in 2001, there was a point of time -- in 2011, there was a point in time in 2009 when watson made a quantum jump in its ability to learn questions. at that point, i remember sitting in my conference room looking at the demonstration. i said, we have something that is going to change the world. it was shocking. >> did people you back then? -- people doubt you? they they agree with you at the time that this would be a watershed moment? >> yes. obviously my research team had hydrogens. they were very focused on the q&a machine. -- had high ambitions. this is not just another computer system. this is a new era of computing. >> but no zombies, no apocalypse, no matrix? >> it's only going to help us

with complex decisions. >> thanks. [applause] >> man, this room is filling up. you are here to hear the investor talk, aren't you? how many on maneuvers are in the room? so all of you. listen up all of you. maybe you can shut your laptops and learn something. these are the folks that are going to give you money. please welcome our guests. big round of applause. [applause] ♪

>> i am so glad to have this great panel. i will roll into things and ask you a hypothetical. say i am this supersmart entrepreneur. i have a great product, i live in maybe austin. everyone is telling me i have to move to san francisco. is that true? money if you're raising from greg croft. >> why is that? >> i think 82% of our investments have been outside silicon valley. we are finding opportunities everywhere. i think there are great growth pockets in austin, new york, ellie, seattle, all over the world. >> i think it depends. if you are based in austin or l.a. and you have a network with

a recruiting advantage, then it might make sense to stay. like there is expertise that you need that doesn't exist in the market you are in, the bay area is quantitatively the best place to found companies. >> is not getting too expensive though? -- isn't it getting too expensive though? being based in san francisco at this point cuts your runway in half. he said, if you live in san mateo, oakland, you can get so much more mileage out of your fundraising. does that change things? spendlong as you don't all of your money on candy bars, you are fine. do think you can found a company anywhere. the challenge is scaling back company to the next level. you can find great individual contributors.

you can find great designers, great product people in any country, in any city, in any town. once you start to scale, let's say it's working and you starts to hit 100 people on the team. you have folks that are not up from theut 2 frontlines. back is difficult the further you are from the bay area. -- that gets difficult the further you are. experiences,t of especially on the consumer side. it only exists here. there are certain cities in the l.a.,-- new york,

increasingly in europe, you are starting to see some of that second generation experience. if you are in tuscaloosa, you are going to have a tough time finding that locally. more importantly, a tough time attracting someone who has that experience to move there. if you're in a global city where people are willing to move, and i think l.a. and new york counts. austin i think is on the edge. it has some livability that makes it attractive to people. it doesn't have the same ecosystem for what people do next. that is something people care about. that is the key question. if you can't attract people who are willing to move to your city, you have to consider moving your headquarters to a place where those people already are. network --ou have a

i would argue l.a. might have the most talented people in the media business out of anywhere in the world. dynamics ofand the that market better than anywhere. if you are starting a digital media company in san francisco, you will be creating -- you would be recruiting executives from l.a. it depends on the network you bring. any great entrepreneur has to be a pied piper and be able to recruit people anywhere, whether you are in reykjavik or in chicago. we have seen great companies in both of those places. that is a huge part of it, to pick a vision and tell a story that attracts people anywhere. >> what about new sections of technology? maybe i will segue into asking what you're looking at that you were not a couple of years ago.

i don't know if there is a virtual reality headquarters. >> john. we just announced this morning, we and disney: an investment in john, the virtual reality company based in palo alto along leadcmc, a chinese investment platform. it's the combination of both technology and innovation that you find here any specific expertise on related to content and distribution that you would find in disney. caa.n the a combination of l.a.-shanghai media landscape married to the technology innovation here. most of the most interesting

technology innovations are happening here. channels areng happening elsewhere. it is the combination of both. funded anything in the last couple of years that you were not have imagined funding a couple of years ago? >> we are looking at a lot of things that, 2-3 years ago, we would not be looking at. the food category known was looking at 3-4 years ago. now it is a $1 trillion plus market. looking at drones. , we have been starting to look at more health-care related things. capital,akes a lot of it's probably not a great fit for us. when things become increasingly consumerized, i think it can make more sense for investment.

that is been challenging for us. i have been hearing a lot more about health care and a core sciencies. holland is a take for you to get up to speed on a new sector? -- how long does it take you? haven't done anything in it, so i can't speak to it. but what are the big sectors that we see? i think that the punch maneuvers --entrepreneurs see opportunities. it's hard to predict the future. it's easier to just observe the present and see what is working, then trying to suspend judgment.

"oh, that is just for kids, oh, that is not going to scale." predicting the future is very difficult. insultingd recently that is given us tremendous growth on the back of social media. it was not tell you going to be a big driver, that this was the right time. none of that made sense. the point is, all we need to do is recognize it, see who the leaders were, and suspended judgment. saying, it's not silly, it's actually a meaningful thing. having the context matching and

say, okay, this fits into a lens of self-expression at has a truth matter of e-mail signature files through aim buddy icons and myspace propos pages. you can start to see that connection. that is where having a sense as to where things are going based on where things have come from is important. here is whereing, the opportunities are going to be. >> we have a slightly different model. we work with investors as partners along with the entrepreneurs that they back. we invest with industries with a long track record. we are the largest health-care investor in the world. biotechnology companies, drug discovery companies, ally, format, etc.. it gives you a different perspective and allows us to be

a strategic partner to companies. we tend to find ourselves not competing with venture firms or earlier stage investors, but rather coming and when we can be a real partner. we got involved with uber and airbnb, we can and because the object newer wanted a partner that was global and had deep industry capabilities. allowednally they have us to scale. we are jumping in the boat and helping them row as opposed to being particularly wise at where things are going to unfold at the earlier stage. >> a lot of people don't realize tbg growth was a large investor in airbnb and uber. cbg asf people think of one of these firms that drives valuations. what do we do here?

it is obviously a problem. are you writing bigger checks? are you writing fewer checks? we are at the beginning of the supply chain of venture capital. for us, this year we have been slower to invest. are yearslysis, there that have lots of great new ideas. last year and the year before were better ideas. ideas are looking for higher valuations. the punch newer -- entrepreneurs that are looking for rich valuations to realize they could be setting themselves on the wrong trajectory for the rest of their life. we try and balance the fact that sometimes the valuation doesn't

matter if you are in on some thing big. you don't want to set the company up for failure because it started out too expensive. this year seems like a great year for companies starting a couple years ago. we have not been super active this year. we are looking for bright ideas at the right valuations. >> agreed. there has been a significant increase in number of violations, but those levels are relatively constant. -- in number of valuations. today, theo pre-money valuations have essentially doubled. that is not healthy for the overall ecosystem. that is where you sit back.

when you look at average exits going back 30 years, 90 plus percent being at $105 million, the math gets tough. there is the reality factor that plays in. we think about that a lot. in the markets where we invest more heavily, some geographies tend to have more valuation discipline, just a lower supply of capital. >> are there any factors that are better deals? certainly people are interested in synthetic biology, something relatively undiscovered. >> we have continued to invest in e-commerce companies at a time when he has been relatively out of favor. never been a has

better time to start a new brand. when you look at the market and look what millennial's are driving, it's a completely different brand message. you can just change your messaging. marketing through a completely different channel. is not television and newspaper. you can't spend at big volume on a lot of each other channels. evenook and instagram are old now. no better time to start a new brand. if they have great economics and a great value proposition and a great brand, we think they are a great opportunity. >> there has never been a better time to be an object newer -- entrepreneur. >> were talking about ipos.

he said there was a private-public confluence. you said that when i change. -- that would not change. at the same time, everyone on this panel wants their public -- their company to go public and julie. are we deluding ourselves? question. a big you are talking about $50 billion that has entered the market to invest that was not here before. so far this year, $7 billion of capital invested from historically public investors. $11 billion invested last year. all of which allowing companies to raise capital at a scale that is unprecedented. which allows for it deferral, the ultimate public reckoning, if you will. >> there was a time when this happened before. >> except even then, if you look ipo's9, 29% of all

doubled in value the first day of trading. 1% of ipo's have done that this year. 29% versus 1%. this isn't a public double today. --bubble today. the bubble is largely private. if the gas gets let out of the bubble, it will be in a slow way. the public market has not been participating. do we see public investors to speak? of course. -- public investors participate? of course. there is liquidity. ipo's have happened at corrective values. >> in this case, is the correction hitting a wall? >> it's a funny thing. if you are a private company and see this correction happening, it's much worse for founders and

employees to have a correction happen as a private company then a public company. you have preferences from all these investors that have put in all this money, who are looking to get made whole before any proceeds go to employees and founders. if you're public, the math changes a little bit. all of the preferences go away. everyone has the same stock. if there are fluctuations, everyone shares in them equally. markets,y private where folks with preferences and those that might you like they have overpaid -- might feel like they have overpaid are going to have much different preferences. now you have better access to additional capital as a public

company. and you have employees and founders in a more beneficial situation if there is a correction. it's interesting there has been this reticence to go public for companies able to do so. [yelling in background] >> sorry, i'm having trouble hearing you. >> we have to get that guy up on the panel. >> i was at a talk last week bout sometalked a companies whose unit economics he didn't understand. he did not imply they were doomed. there are 150 companies evaluated at more than $150 billion. are there any names that you can offer? a company that you don't know what its fate will be, but you don't understand it's economics. >> i will not name a name,

because that would be indiscreet. but we have to try to make investments where 100% of the investments work given the scale of the checks we are providing. if you look back at 1992 and you made 100 investments and it was in the tech sector, unit of making a 20% return on your capital by the time you exited 10 years later in 2002. but it would be-concentrated in a microsoft and intel. if you could invest across a basket of unicorns, as eileen contended in 2013, you would do just fine. you have a few of these companies that do extraordinarily well, but there will be a lot of roadkill and companies that don't survive. their unit economics don't work.

ultimately in rising markets, people don't ask the hard questions about cash flow generation, unit you know, leverage of scale, etc.. they focus on revenue. and we have all learned this lesson before. ultimately, they have to get to a place to generate cash, and valuations are ultimately about cash flow. i have about 20 more questions, and we have to get going. thank you so much. [applause] ♪ anchor: and we're going to go to lunch. just kidding. you guys are all here. what are you excited about? i'm going to do this part of my job and relish it. please welcome to the stage

snoop dogg. ♪ snoop: what's up? anchor: we go way back. wherever you like. actually, you are going to sit in the middle. you are here. snoop yes. anchor: it is all happening. live, and in the flesh. you have been investing a while now. first of all, if i invest, it has to be something that is fun and amazing. i associate myself with it, and that is what i am looking for when i do invest. of your in terms

investing, are you reinvesting with some plan in mind? are you seeking something out? in the beginning, it was investing for creativity, and then there were things that i like. for example, the cannabis industry. i saw things in that field and saw that there are things that were missing, so i thought the best thing i could do was to develop my own system and get in on that side. something, and you might want to tell us a little bit about that? snoop: it is called mary jane. helpn watch a video to show. if you do not mind, we would like to show you a little clip. [video clip] snoop: in the world, mainstream

cannabis. let's modernize and communicate. people all around the world use cannabis, and there is a global scale. in 2015, cannabis was named the fastest growing industry in the united states, set to reach $11 million per year by 2019. -- $11 billion per year by 2019. fashion, global music and sport, and today, mary jane cannabis. mary jane cannabis 2.0. for thew generation cannabis culture for all. jane has the global cultivation of cannabis.

jane for everyone as the center of cannabis lifestyle. jane haslly, merry tools to bring together consumers and businesses in a streamlined fashion online. it is a revolutionary way to discover all products for a new era of consumption, and it is for main street to find premier locations. has that cannabis is part of our daily lives. merry jane is at the helm of the move. [end video clip] [applause] a whole new world ♪ ok. what kind of partnerships are we going to see or the network effects are we going to see?

think this is probably the priority we are doing in the cannabis space. obviously, we have been at the helm of this movement globally, and the time is just right. we saw, and as you mentioned, a huge tranche with the way that the conversation is around cannabis and what is provided for the community, and for those who are touched by cannabis and want to learn more about it. it just was not available, and it was not done authentically. there was a certain sophisticated standard of content, which is really our expertise, along with some of the partners that we have in this space, like my good buddy joining us in the content space. he will provide us with

some content, and you can find some great content from mr. ch ung on the merry jane site. the and then there was miley cyrus helping brand merry jane during one of the gets, so it is getting the top-tier theuencers and sharing modern situation around cannabis. jordan: this seems like an obvious question, but, snoop, why do you like cannabis so much? i mean, it has been such a huge part of your brand. i want to know why? snoop: for one, i enjoyed it for

medical reasons. [laughter] jordan: trouble sleeping? snoop: every time i have been around it, i have seen beautiful things happen. love, peace, happiness, and i have always been an advocate for it. it is more educational and and it gives off more than just me trying to smoke and to give you an explanation. ityou want to learn about come we provide you all of the information. it is an encyclopedia to the cannabis world. jordan: well, what kind of we expect to see? i know you said there would be celebrities, but there are editorials, videographers. what you expect from this kind of content?

ted: there really is a diverse range and we are talking about cultural content. we have a cooking show, food, and a bunch of series coming out, and to kind of expand further, the great thing about the cannabis industry beyond its medical benefits and social benefit, the job creation -- we're talking about a business that is going to be in the $19 billion range, and that is just in a few years. there was over a billion dollars in colorado last year alone. and then the board of education. where else in america are you going to see a cultural revolution like with cannabis, and that is a beautiful thing about what we are doing. it is a movement, and the integration into what and as far as,

series, we do have one thing to show you really quick. opportunity to meet captain mike owens and a sergeant from a charity that helps veterans reintegrate into is aorkforce, and this very cool, one on one interview series, where people enjoy cannabis and share a new life experience that they have, so if we could roll that, we will show everyone. [video clip] mike: i am mike, a former marine captain. marine am a former infantry sergeant. ♪