A conversation with Ginni Rometty, Chairman, President and Chief Executive Officer of IBM

Well, certainly the largest is a part of the reasons I think I came to Davos this time. And what we talked about at the time, that there were three big waves of technology which make this a different point in time than ever before. So when people say, 'Oh, I've seen this before, this has happened before,' not really true. And you've got the rise of cloud computing, which in this sense the reason it's relevant is the ubiquitous nature that you can get things everywhere. The second thing was the rise of data, and we'll come back, that'll be at the heart of this, the rise of data. And then of course you have mobility. But I always have added another one. I always say but underscored with security, which we'll talk a little bit about at some point, because security and privacy are things that will derail one way or another here. But those were the three, they came together. And I think, you know, when you said this is my first time back at Davos in a long time, and part of I was sharing with someone yesterday, the reason I wanted to come back: the subject is responsible leadership. And of course what jumps to all our minds are things like health care, education, and intersecting with those is this topic of artificial intelligence. And we can talk about why we call it cognitive, but artificial intelligence. And I do see it as a solution to much of this. But back to those waves, you have so much information, and this is what got us started on the journey long ago, so much that if you don't do something, your brain, us as humans, cognitively can't deal with all of that. A doctor, 8,000 papers a day, your doctor cannot be current, it's impossible. And so what would you do? And today, every system ever built has been programmed. I mean, your phones, no matter what you use in technology to date, it's been programmed in some way. These systems, you can't program for all of the different if-then what could possibly be. So you need a system that could understand all this data, could reason over it, and could learn, which means they do become more powerful with time. So that to me, this intersects this idea of what all these big issues are in the world right now, and it is in part a solution to it. So welcome back to Davos, that's why.

So if you look at even our history, there's a new movie out right now called Hidden Figures, and it's about John Glenn orbiting the moon and the earth. And in that movie, though, it's talking about the IBM mainframe, it's the first one, it's the first time it's being used here. And that is this programmable era, and it was with math, the answers were deterministic. And what's out there, and so when we did chess, those are deterministic. You can, with a powerful enough system fast enough, you can calculate every move. And dissemination falls very true with these games. You see people having out on TV and the like playing other supercomputers, playing these games, they're mathematically determined. The difference with what we did in Jeopardy is it's open domain. And so as I say, what these systems do, artificial intelligence that understand, reason, or they can deal in gray areas. But even then, it was an open domain, a question and answer. And so if someone asks you a question about, 'Well, where is that chick?' Okay, is that slang for a young girl or is that a little chicken? And for these, it doesn't have context of what it is they're talking about. And so Jeopardy was a game show where the answer was given and you have to determine what the question is. So the ability to not be able to phone home to anyone, for Watson to have been fed lots of literature, and so he's read lots of literature, and then to get an answer and then be able to back up and parse into what could have been the question, that's not search. I keep saying this is not keyword search, this is actually understanding context. So that was a first foray into Jeopardy, and it was with Watson to prove that. But we certainly then moved on, because that was really never just about it. This was a mission for us. It started long ago, so we're probably 15 years now into this. And when we said, 'Hey, the world is going to have all this data, you're not going to be able to deal with it, and it's not all text, things don't all fit in rows and columns, these will be pictures, images, tweets, sound, sensors, video.' So something has to understand all that, that's unstructured data, of which the world is 80% of that kind of information. So that embarked us on the journey. We picked this milestone to do a game show, which a year in advance, our researchers will tell you they weren't ready. At a year in advance, we were like, 'Whoa, okay, this show does air on this day, you're going to work Watson.' That loss would have been humiliating. But the real point was to say, okay, to show the art of the possible. And to me, this is what's much more important and what has moved on. I think these systems have the opportunity to do for some of what have been the world's most unsolvable problems, to find solutions to them, particularly health care. And that was the first big one that we started on. But it's health care, it's education, but it's retail, it's your everyday life. So it is from the everyday to the unsolvable, how I like to think of it. And if I give you sort of two examples on each end, what you can do. So in health care, right or wrong, we started with the hardest thing. We started with oncology. So what we did with Watson, Watson had been fed all textbooks, all journals, about 18 million papers. And then we started training with some of the best cancer centers in the world, because these systems do have to be trained. You don't just pour all the information in and say, 'You're a doctor.' And by the way, I want to be really clear, and I've been so emphatic, I want to come back to a policy letter I issued to our company yesterday. These are technologies to augment human intelligence. In fact, I actually don't like the word AI, because cognitive is much more than AI. And so AI says replacement of people, it's carried some baggage with it, and that is not what we're talking about. By and large, we see a world where this is a partnership between man and machine, and that this is in fact going to make us better and allow us to do what the human condition is best able to do. And I have to tell you, we have seen this play out that way with every profession we've worked through. You watch it become not a thing, it's a very two-way relationship, and it is a tool that helps and it makes you as a professional do a better job. So we started, as I started, on augmenting and with the doctors training on oncology. So I fast forward to where we are today. Watson has learned. We started with breast, lung, colon, the hard body cancers. By the end of this year, he'll have trained on what causes 80% of cancer in the world, 80% of the cases. And explain how you train. What goes on? As I said, you don't just throw the data into a machine. No. So first, very early on, had to learn the language, which is true in every profession. Think about whatever job you have in the room, every profession's got its language and what it means. You take it for granted now that you know it. So we have to teach it the language of medicine, and then what's associated. And so as you give more information, ask questions, check answers, it right and wrong, it begins to learn over time. So fed the information from the journals and textbooks, then in fact the training went on, tens of thousands of hours with these doctors. Then we took medical records where you knew the outcomes, and you would give him the input of the EMR, you would give him the x-rays, diagnostics. Learn, did you get it right or not? So back and forth, and you would improve the accuracy, improve the accuracy. And so you get to a point where you could diagnose and you could offer treatments. And around different kinds, you might put limiters. You know, if I find a cancer patient of a certain age, I still want to have children, there are certain things I don't want to take as medicine and the like. And so you could put different in. And what we learned, and this will be true with every profession, people don't want a black box. They do want to understand, this is assisting me. So give me the percent confidence of this diagnosis, give me the sources of data that fed in primarily to do this. So where we're at now, we are rolling out in hospitals across India, China, Thailand, Finland, Italy, three more, and of course in the United States. We started as an oncology advisor, then we moved on to clinical trial matching. So anyone with breast cancer at Mayo Clinic, clinical trial matching, which is very difficult to do otherwise. And then to genomic sequencing, really progressing precision medicine. So Watson has been taught by the top 20 genomic centers. And if you are a late stage cancer particularly, you have your whole genome sequenced. Now with Illumina and with Quest Diagnostics, they run it through Watson and you get the different sets of possible both diagnosis and treatments that come out.

Well, perhaps yes and yes. Right, because what Watson today, and I'll use Watson as our example, it's all our experience. We were the first one. We've got what I call the Watson, it's the AI platform for business. We will touch a billion people this year, not us directly, people using Watson are touching a billion people. And so we have a lot of lessons and learnings on this topic. Now, if you go out, and you all can because it's an open platform, people are building anything on this. And I think I don't even know, we see them because you go out. So one of them is the personality insights and the tone analyzer. And with David, it can tell me how many words, and personality insights, it's maybe not very many words. You're about 150, I think that's equivalent to a tweet, you know, 140, about 150 words. So 250 words, you can get a pretty decent, we can do you, which I should have done you ahead of time actually. I got to do that for the next time I see you. Oh no, I can't wait. So I didn't think of this, and I got plenty of his writings to put in there too, so that's not going to be a problem. But you get percent insight and your tone. In fact, I'll get sometimes notes from folks at work and they'll put their notes through the tone analyzer first before they, you know, get it. Yeah, like, 'We just was telling, were you trying to write that in?' So yes, so that is true, you can do that. And I think these are choices we're going to make about trading off some. In some cases, you don't know you're trading it, which I think you have to be aware of when you're trading off information and how it's used. And you'll make some of those trade-offs for convenience. But I do think we're going to enter a world that has to be in your hands whether or not you want to live and work that way. Some people will trade that off for convenience, they do today. You trade off your location because you want to know where a restaurant is and this and that, or directions. So you're making a conscious decision, 'I don't want to share my location, so I don't do that.' But you'll do that, and that is the world. You can do those things, but I don't necessarily find that bad. And I do think it will enhance a lot. So you said, 'Could I make something better?' I can't run Fruitvale. Last time I seen you, we had done this. We did some work with a music producer called Alex da Kid. So unless you're into current contemporary kind of music like this, and what we did, he took, and we worked with him, but it just shows you how it can improve the creative take, but to improve what humans can do creatively. He's a world-renowned record producer, having done people like Rihanna, Beyonce, a contemporary artist of our time. And took all of the, we took all of the famous top sort of 100 songs of every year for a great number of years, fed them in. We almost analyzed all the conversation of the world because we took all of the Twittersphere, everything for last five years, analyzed it. He got a lot of information about mood, what people react to, the music, whether they liked it. And to make a long story short, about two months ago he put out a song and it went to number one on Spotify and on iTunes in the shortest period anything has ever moved. And so did he get lucky, or was this the combination of these two things? We've seen it, you and I talked about a movie trailer. Yeah, we've done it to enhance that. And you've seen the same thing. In fact, in that case, the funny story was the movie trailer was more famous, more watched than the movie. And the reason this is interesting is movie trailers was always regarded as a great editing art that you had to, a great editor had to look at a movie and figure out what were the most, you know, the moments that would most likely make people want to watch the movie. And Watson did it essentially better. So all alone, we're going to talk about jobs. Does make everyone wonder, if the computer can do all this, what do human beings do?

Yeah, so this is what our thousands and thousands of lessons and experiences taught us, though. And it is a really important part of being here, that it's why it's augmented intelligence. And in all those cases, the person played a very important part in things that the system could not recreate. And so I view this as an era, this will play out over decades to come in front of us. But there are some things that are really important. And if I might, I want to share main tenants of a policy letter I sent to my IBMers yesterday. So I have almost 400,000 IBMers, and the last time a policy letter was written was a policy letter that said, you know, if you had genomic testing, we would not use that for any form of information or discrimination. They are very rare. I wrote one yesterday on the principles for the cognitive era, or think of it as the principles for AI if that's better. And it gets to the heart of this. And I said there are three things, because history has taught us many things, and us at 105 years old, that when you introduce powerful technologies into this world, you have a responsibility that they are introduced in the right way. And you can guide their adoption and you can guide how they are used. I go back to the 1960s when IBM first came out with it, it was really that big massive programmable system. To this day, reinvented many times, but that gave rise. Our role was to teach the world about how that could be used. And that, you might say, automated many of the back offices of the world. That gave rise, and we played a big role in education to computer science. There was no computer science at the time taught in universities. That was really, I go back and look at pictures from that time, they're all pictures of classrooms and teaching. And so I fast-forward now. What are the three principles that I shared with the IBMers to guide our work, what we do, what we believe the world to do, the industry should do, what we will devote ourselves to? So that's very important and powerful. The first one is these technologies, we will be clear of their purpose. And our belief is these purpose is in service of mankind. They are in service of humans. They are here to extend what you and I can do and to extend the human capability. We do, and we debated these. We do not believe either in principle or actually even in the state of science today that these will be self-aware or conscious. And that's not what we're advocating for. And so that, if you put simply, the human is in control, ultimate control here of what happens. So that's the first principle. And we really, David Kenny's on my team, we debated these. Number one, purpose is to extend what humans do in service of. Second big principle is the word transparency. I think time has taught us this. So we need to be transparent with everyone when, and we believe others should, when are you using artificial intelligence? Tell people. When is that answer come with this? The second most important is how are these systems trained? Who trained them and what data was used to train them? So would you, if a doctor knew that the top 20 best cancer institutes trained that, what's the likelihood he's going to listen more than, 'Well, this came from somewhere, this came from scraping the web.' I mean, no, that's not how an underwriter works, it's not how a doctor works, a teacher works. So you need to know where and who taught it. That's what we're doing. Financial risk systems now, we bought the world's most renowned company for doing financial risk and they're doing the training. So you need to be clear. And the other part of transparency is on business model. This is quite, to me, this is a wedge issue and quite concerning, meaning all data and algorithms should not be concentrated in one company. This is not a good thing. So with a business model, as a company, as an example, those of us in the room of companies, you've got accumulated decades of knowledge. Do not turn it over. You should know what you turn over. And when these algorithms are trained, that insight belongs to you. That you are not training someone's data to help your competitor. And that's how some of the other systems work out there today. So transparency is the second big principle. And the third principle is around skills, which we'll end on that topic. The importance of guiding and building the skills in the world to effectively use this technology, safely use this technology, put it in the right service, and be sure the right jobs are created and that you reskill where that's required. So those three principles of purpose, transparency, and skills are something I am going to adopt in my company and I hope people globally do. It's something I think whether it's government, company, academia, it's really important because we're just at the beginnings of this era.