How many people here can have a feel for what AI really is? We’ll talk a little bit about what that is. But I kind of want to turn the clock back for a few moments. So let’s wind back and go to June 29, 2007. Does this date ring a bell for anybody?
Iphone. Good. It’s the launch date of the iPhone. Think about it. The iPhone is only so many years old. Does it seem like it’s that young? Think about how much smartphones have changed our lives, both personally and professionally. It’s created this whole – what they call – app economy. Companies like Uber or Snapchat can’t exist without smartphones. Right? It’s something that’s really revolutionized the way we do work, the way we live our lives. And I think we’re on the same precipice with artificial intelligence.
Now, I know there’s a lot of press, a lot of debate out there about what’s going to happen. What does this mean? I think there’s actually a real opportunity here to change the game across the board, particularly in social good. The question is, how do we actually make that happen? So I’m going to share some secrets with you about that. But first, we’ve got to figure out what AI really is. So in a nutshell, I really don’t say I like to call it cognitive computing, because as far as I know, there are no machines out there that are self-aware, so thinking.
So for cognitive computing, we’re really building machines that mimic the way humans think. So it’s not your traditional type of computer, right? It’s not going to execute a set of instructions. It’s not following a path or exception path. It is actually taking information in using its experience, its knowledge to try and formulate an answer. And in that way, there are more than search engines, right? If you ask a search engine and something that, you know, we can’t just look up, it doesn’t know what to do. But with an AI type of machine, you can actually ask questions and actually try and solve problems you don’t have answers to, which is why we can do things like ask AI machines to come up with new target proteins in the fight against cancer. So what is I, I think there are three special things to it. The first is really this concept of machine learning that the more it does, the better it gets. I believe there was a reference made to Malcolm Gladwell’s 10,000 hours – any time an AI tries to do something, it’s like a three year old kid. It doesn’t really know that much, but you have to teach it some basics. And then you let it try.
You try to correct it, you tell it, “Hey, this was right. This was wrong. It should have been this. This is partially right. A better answer would have been this.” And it learns and it keeps doing. You keep teaching, it learns. But to go from three year old to PhD, it only needs a matter of a few weeks. Think about that for a second. That’s very fast learning. Second is this ability to process natural language. And I can’t tell you how difficult it is for machines to handle this. Think about the way we talk. Do we talk like we normally should? No. He’s a lot of slang, jargon, idiom. A lot of our speeches, body language. There’s a lot of contextual understanding. If I told you guys that I’m feeling blue because it’s raining cats and dogs, do you understand what I’m trying to say? If I told the machine, what’s it going to think? It’s going to be like, “Neil is somehow now the color of blue because small animals are falling from the sky.” That does not compute. Right? But this is the special sauce for AI and that it actually does understand that. The third piece is this ability to actually interact with this. As a human being, we can actually have a conversation with the machine.
For better or worse, we’ve actually been trained to use keywords. You think about how we use Google or use Alexa, what happens, right? If you say, “Hey, Alexa, turn off the lights, or sorry, hey, Alexa, turn on the lights,” it’s going to turn the lights on. But if I go, “Hey, Alexa, it’s dark in here,” what does it do? Nothing. Doesn’t understand the AI machine. If I were to say, “Hey, it’s dark in here,” it would actually probably turn the lights on because it understands the context. What I’m trying to say (is) with these three pieces kind of stitched together, this is what we really think of as artificial intelligence. And what we get out of that is actually this ability to do new things, right? We’re unlocking new capabilities we’ve never really had before. And as a result, we have a whole new way of thinking, right? We’re no longer looking at machines as a way just to automate something, do it faster, do it cheaper or whatever. We can actually now leverage AI to help us solve more complex problems because to be perfectly honest, even the smartest humans, we can only juggle 7 to 12 variables in our heads at a time. You think about all the data we’re collecting, all the things that might influence a particular decision or a search for a cure.
We’re trying to find the best way to plant our crops. There’s probably millions of variables involved. So we have this opportunity to actually build AI powered tools to actually help us get better yields. To where is this actually happening? It’s across the board. Today, I can tell you that AI is being used in pretty much every sector, every industry out there in some fashion, from health care to education to media entertainment to life sciences to legal services to financial services. There’s actually not an area that hasn’t been touched out there. If I asked you guys, “Do you use any AI tools in your lives?” How many people would raise your hands? So I’m seeing maybe about 40%. I bet you’ve used tools without even realizing it. If you ever use one of those chat bots out there now, about a third of them are actually powered by some AI machine. You’re not actually chatting with a real person anymore. If anybody wants to admit that they use online dating tools, they’ve shown, I think, on Tinder, that over half the profiles now are actually bots. Yeah. The bots are so sophisticated they can carry a conversation for levels deep before you might realize it’s not a real person. Something to think about. I know, right? With this, this is really the value we get from AI. First and foremost is this ability for new insights.
One of the first things that we actually did with IBM Watson was go into health care and we try to do cancer research, oncology, these types of things. But because machines can learn so quickly that it’s become very adept. There’s actually a case recently where there was a Japanese woman that fell ill so her normal doctor couldn’t figure out what’s going on, brought up some specialists. They tried some things, wasn’t working. And after several months, one doctor says, “You know what, let’s give Watson the shot.” So they brought Watson in. Watson looked at all the work the doctors did, all the medical tests, looked at genetic sequencing, spent about, I think, 20 minutes with the patient and then diagnosed and said, I think she has two rare forms of leukemia. You know, people are like, I don’t know if that’s really true, but they tested for it and she tested positive. And now that you knew what you actually had, you were actually able to start a proper treatment plan, right? So think about that. We can actually build tools using AI where if you were to go see your doctor in real time, they could have a little AI assistant with them to help. Help with it, help the doctor with the diagnosis and the treatment. So you come in, the AI could be watching you, listening with the doctor looking at the symptoms that you’re displaying, looking at symptoms you may not be displaying.
Looking at your medical history, medications, all these things to help arrive at what’s probably ailing you. And then that’s what we’re looking at. Second is this ability to actually help us reduce bias. Right. If you think about some things, obviously we’re shaped by our experiences, that sometimes you block out potential good ideas because we don’t think they’re feasible. For example, how many guys would put peanut butter on a bacon cheeseburger? It doesn’t sound appealing. Right. But it might be delicious. I don’t know. But because if you don’t think it’s appealing, we don’t try.
One of the things that we actually did was we created an app called Chef Watson. And what we did was Watson has the ability to actually come up with original food recipes. And one of the things we asked for was a healthy barbecue sauce. And Watson came back and said, “Well, what do you mean by healthy? Remember, interactive?” And so that’s a good question. How healthy? Let’s go with low calorie, high fiber. Watson’s all right, comes back. Recipe has, like, butternut squash terman. Like, stuff you’d never find in barbecue sauce. And we’re like, okay, it doesn’t work. I mean, what does the machine know about eating? Right. But it actually was tasty and it was healthy. Right. And how could Watson come up with something like this? If you think about how you’re going to teach a person to cook, what would you do? You would probably share recipes, maybe watch YouTube videos.
Taste testing machines don’t eat. Machines have no sense of flavor. So how do you teach the machine? Well, we taught Watson chemistry. We taught Watson the chemical combinations that produce flavors humans, produce the nutrition ingredients that we need. And because of that, Watson was able to come up with all kinds of original recipes like the Chocolate Austin Burrito or the Vietnamese apple kabob. But more importantly, you could extend that out further. You can then find good substitutes to some recipes for allergies, like if you have dietary restrictions or you’re lactose intolerant or you have a gluten allergy, or you can even take a step further and say, “Well, what if I want to only make recipes that are locally sourced?” Then you can watch and you can locate what local ingredients are and help you make that dish or come up with new dishes. Take that a step further. I didn’t say like, well, based on where I am, what kind of food could I actually grow? Right. What demand might be and what kind of crop yields might I need? So there’s a lot of power and a lot of synergy in what we can do.
Lastly, what I want to call out is this ability to actually understand emotion. Now there are things that as human beings were much better than machines at, but there are things that machines are better than us at. And some of them we have a tough time acknowledging. If I were to ask you guys, could a machine actually understand human emotion? Who actually thinks that’s true? That’s a surprising number of hands, actually, because most people will find that hard to believe. But machines can understand human emotion. We can teach them about body language, tone of voice, word choice, so they can actually infer the emotional state of a person and actually respond to that person accordingly. So if you had an AI powered robot somewhere, and it was having a conversation with somebody and detected that person as sad, they might try to cheer them up if in the middle of conversation the person turns angry, might try to placate them. We’ve actually learned that in some cases, machines are actually better at reading emotions than we are. Is that surprising? If you think about it, read body language. How good are we at that? On our face there’s over 2000 tells. How many can we actually look for in real time? Machines can see all 2000 in real time. What’s more surprising is that we’ve actually found that people are much more open and honest to a machine than they are to humans.
Think about behavioral health like a therapist or psychologist or your doctor or even like your financial analyst or something. People who have been working with the same person for 20 years, have a trusting relationship, are more likely to tell a machine something than tell that person. And we were surprised by this. And what we learned was people, when they talk to the machine, they feel like they’re not going to get judged. They’re not going to get embarrassed by saying something. And because of that, they’re willing to share more information in that regard. I mean, she can actually learn more about the person and actually help us come with a much better personalized solution. That’s why it’s the real power of AI. It’s not really about finding the right solutions anymore. It’s about the best solution for any given situation. You know, a real simple example, how many times you just like hanging around with some friends and somebody says, let’s go to lunch and like, great, where do you want to go? I don’t care. Where do you want to go? I don’t know. Right. There are four of you, right? You had a little AI system to help you find the right restaurant. Might know that, okay, this guy over here doesn’t like Japanese food. This guy doesn’t like driving very far. This guy just likes paying a lot of money, you know? Do all the matchmaking and figure out what the right restaurant should be.
You swap one of the friends out with somebody else. It’ll do it. Come up with a whole different answer. Right. Different dynamic. And that’s the real power of AI, that we can look at a specific type of problem, a specific type of situation, and come up with an optimal solution for that specific situation. So a lot of the examples you guys have probably seen are very commercial. And you know, I get that and everybody wants to make money, but I think there’s a lot of good that can also be done and there’s no reason why we can’t do both. I think there are plenty of opportunities for commercial and social good solutions out there. And one of the things I’ve been lucky enough to work on with several groups, NGOs and the United Nations is trying to see how we can use AI to fulfill the Sustainable Development Goals. So I don’t know how familiar you are with them. There’s 17 of them. But the United Nations has estimated the resources we currently have today versus the resources we’re going to need to actually try and fulfill these goals. The minimum shortfall is about $5 trillion. It might be as high as $9 trillion. And then we live now, unfortunately, at a time where there’s degrading emphasis on some of these areas and cutbacks in funding.
So what can we actually do about it? And so a lot of organizations are now looking to say, how can we use technology to try and make up this shortfall? This is not really using trying to do okay, let’s do more with less. But how can we do more and do more better with what? With technology we have out there. And so if you think about biodiversity, what can we do with AI? Well, we talked a little bit about how we can use AI to help improve crop yields, analyze soil content climate, whether even daily sunlight exposure or cloud coverage, humidity, moisture. Find out what the optimal crops actually produce, help reduce the amount of water to grow those crops. We can look at things that might impact the ecosystem as well. You think about what’s happening with bees today. Most people think, well, is that going to interrupt the honey supply? But bees actually do a lot more in that. They help cross-pollinate plants. We can use AI because their ability to assess so many different data points kind of see that whole picture can help us kind of weave what the overall impact of the ecosystem is going to be. So we have a much better picture of the problem we’re facing. So you might be asking yourself, I’m hearing a lot of great things.
Where do I begin? I would say begin at the beginning. Start with the problem in mind. Too often I see these days as people come up with some cool idea, they build something and then they’re like, What do I do with this solution? I have to try and find a problem to wedge it into. That’s not an effective use of I think about something that you want to solve or something you’ve always wanted to do but never been able to do. Could I help you with that? And if you actually want to figure out how to do that, the secret sauce is two steps. One, forget everything you already know. Right. We’re so locked into our mode of thinking that AI actually requires us to think differently. If you think about driving, how do we learn how to drive? We do it very visually. And so we’ve created self-driving cars. A lot of it’s based on visual type of information, like in proximity sensors. But what if you could hear a child running across the street before you see it or see him? Right. We have the technology that can actually enable cars to do that. So we have to think differently to get the full value out of artificial intelligence. So the first step is to kind of forget what we already know. The second step is to actually think differently and figure out what we can do?
If you want to hire somebody, what are you looking for? Are they qualified? Are they a cultural fit? Well, qualifications we can test for, right? We can do interviews and other types of things. But how do you know someone’s a fit? Especially when people don’t act like their normal selves during an interview. Right. It’s kind of like online dating, right? These people don’t go dirt bike racing and hang out with tigers every day. But with AI we actually have the ability to generate psychographic profiles right there, AI tools out there that know psychology. So based on your resume, based on your LinkedIn profile, your Facebook posts, your tweets, what public information might be out there, they can actually generate your personality profile and say, well, this person or not really will be a fit for your team, for your corporate culture. With that, it makes it a lot simpler. So now suddenly we’re solving a problem we didn’t think we could solve before. And we’re trying to do that in social good. So there’s a lot of work actually going on today in sustainability. So there’s a lot of things going on in agriculture and biometrics and energy usage. We are trying to make the place a better world. And one of the ways we’re doing that is with an initiative called Project Lucy, where a bunch of governments, NGOs, other organizations, including IBM, are committing time, resources and money to help enhance the infrastructure in Africa.
So if you look at something like agriculture, where some of these farmers don’t have access to a lot of water for irrigation, you can actually help them do that and help them grow their crops in less water or in health care where there’s about, I think, one doctor for every 2000 people. What can you do? Well, you can create, again, a little AI assistant, a little tablet. You can get villages so that if someone has a problem, the villager can use that to help diagnose and treat a person. So we actually have this ability to help fill in some of these gaps, can actually help solve all these problems with less resources, less time, and overall less money. So the question really becomes then how do we actually leverage AI? Obviously, I’m only scratching the surface and I’ve got to invest a little time in research and understanding. But, you know, if you look at the iPhone, nobody would have predicted what would happen over the next ten years. People ask me all the time, Where do you see AI these next ten years? And the honest answer is, I have no idea. I’ve seen so many amazing ideas already. There’s the guiding eyes, using AI to train scene AI dogs and do matchmaking based on the dog’s personality and human’s personality to find the optimal fit. I never would have guessed that a couple of years ago. The only thing I can tell you about where we’ll be with AI in ten years is I guarantee you that at least 90% of every product or service out there will have some component of AI to it. And that’s why I have a call to action to you guys.
What’s the major pain point in your life or your job or what problem do you want to try and solve? You know, think about how you can solve it. Forget about people, process technology, think pie in the sky and then see if I could actually help you do that. Right. I’ve given tons of ideas away because I get to date with people, figure out new solutions, especially in social good. But I do so much of that work that I kind of wish I had my own AI brainstorming assistant. So think about what might be able to help you guys. And that’s my call to action. My challenge to you is how would you use AI? Because if you’re not already thinking about it, you’re actually two years behind the curve.