Generative AI: Its Rise and Potential for Society

Generative AI: Its Rise and Potential for Society

Show Video

Malcolm Gladwell: Alright, welcome everybody. You guys excited? Hello, hello. Welcome to Smart Talks  with IBM, a podcast from Pushkin Industries,   iHeartRadio and IBM. I’m Malcolm Gladwell. This season, we’re continuing our conversations   with New Creators— visionaries who are creatively  applying technology in business to drive change,   but with a focus on the transformative power  of artificial intelligence and what it means   to leverage AI as a game-changing  multiplier for your business. 

Today’s episode is a bit different than usual.  I was recently joined onstage by Darío Gil for   a conversation in front of a live audience at  the iHeartMedia headquarters in Manhattan. Darío   is the Senior Vice President and Director  of IBM Research—one of the world’s largest   and most influential corporate research labs. We discussed the rise of generative AI and what   it means for business and society. He also  explained how organizations that leverage   AI to create value will dominate in the near  future. Okay, let’s get to the conversation. 

Malcolm Gladwell: Hello, everyone. Welcome. And I'm  here with Dr. Darío Gil. And I wanted to say,   before we get started—this is something I said  backstage: that I feel very guilty today because,   you're the arguably one of the most important  figures in AI research in the world, and we   have taken you away from your job for a morning. It's like if, you know, Oppenheimer's wife in   1944 said, “Let's go and have a little getaway in  the Bahamas.” It's that kind of thing. You know,   what do you say to your wife? “I  can't. We have got to work on this   thing I can't tell you about.” She's like, “Get me out of Los Alamos.” “No.” So I do feel guilty.  

Um, we've set back research by about four hours. but I wanted to—you've been up with that, with   IBM, for 20— Dario: Years. Twenty years this summer.  Malcolm: So—and how old were you when you—not  to give away your age, but you were how old   when you started? Dario: I was 28.  Malcolm: Okay. So I want to go  back to your 28-year-old self. Now,  

if I asked you about artificial intelligence,  I asked 28-year-old Darío, “What does the   future hold for AI? How quickly will this new  technology transform our world?” et cetera,   et cetera, what would 28-year-old Darío have said? Dario: Well, I think the first thing is that even   though AI as a field has been with us for a  long time—since the mid-1950s—at that time,   “AI” was not a very polite word to say,  meaning within the scientific community.  People didn't use, sort of, that term. They  would have said things like, you know, maybe, “I   do things related to machine learning,” right? Or  “statistical techniques, in terms of classifiers,”   and so on. But AI had a mixed reputation, right?  It had gone through different cycles of hype and,   it's also had moments of a lot of negativity  towards it because of lack of success.  Um—and so I think that that will be the first  thing. We'd probably say, like, AI is like— what  

is that? Like, you know, respectable scientists  are not working on AI defined as such. And that   really changed over the last 15 years only, right?  I would say with the advent of deep learning,   over the last decade, is when that reentered  again the lexicon of saying “AI,” and that   that was a legitimate thing, to work on. So I would say that that's the first thing—I   think we would have noticed a  contrast 20 years ago. Yeah. Malcolm: So at what point in your 20-year  tenure at IBM would you say you kind of   snapped into the present kind of “wow” mode? Dario: I would say, in, maybe the late 2000s.   When IBM was working on the Jeopardy! project,  and just seeing the demonstrations of what could   be done in question- answering; it— Malcolm: Literally, Jeopardy! is this   crucial moment in the history of AI. Dario: You know, there had been a long   and wonderful history, inside IBM on  AI. So, for example, in terms of, like,  

these grand challenges at the very  beginning of the field’s founding,   which is this famous Dartmouth conference  that, actually, IBM sponsored, to create,   there was an IBMer there called Nathaniel  Rochester, and there were a few others who,   right after that—they started thinking  about demonstrations of this field.  And then, for example, they created the first,  game to play checkers and to demonstrate that you   could do machine learning on that. Obviously, we  saw later in the ’90s, like chess, that was a very   famous example of that. That was Deep Blue. With  Deep Blue, right? And, playing with Kasparov.  And then—but I think the moment that  was really—those other ones felt like,   kind of like brute force, anticipating sort of  like moves ahead. But this aspect of dealing with   language and question-answering felt different.  And I think for us internally and many others,  

was when—a moment of saying like, wow,  you know, what are the possibilities here?  And then soon after that, connected to the sort of  advancements in computing and with deep learning,   the last decade has just been an all-out,  you know, sort of like front of advancements,   and that—and I just continue to be  more and more impressed. And the last   few years have been remarkable, too. Yeah. Malcolm: I'm going to ask you three quick conceptual questions before we dig into it. Just so I sort of get a—we all get a   feel for the shape of AI. Question number one  is, where are we in the evolution of this? So,   the obvious que—we, we all suddenly are aware  of it, we're talking about it. What—can you give   us an analogy about where we are in the kind  of likely evolution of this as a Technology? Dario: So I think we're on a significant  inflection point. That, it feels like  

the equivalent of the first browsers when they  appeared, and people imagined the possibilities   of the internet—or more, imagined experiencing  the internet. The internet had been around,   right, for quite a few decades. AI  has been around for many decades.  I think the moment we find ourselves in is  that people can touch it, and they can— before,   there were AI systems that were like behind  the scenes, like your search results,   or translation systems. But they didn't have the  experience of like, this is what it feels like to   interact with this thing. So, that's what I mean. I think maybe that analogy of the browser is  

appropriate because it's—all of a  sudden it's like, whoa, you know,   there's this network of machines, and content can  be distributed, and everybody can self-publish.   And there was a moment that—we all remember  that. And I think that that is what the world   has experienced over the last nine months. So, and—but fundamentally, also what is  

important is that this moment is where the ease  of—the number of people that can build and use AI   has skyrocketed. So over the last decade,  technology firms that had large research   teams could build AI that worked really well,  honestly. But when you went down into, say, hey,   can everybody use it? Can a data-science team in  a bank, go and develop these applications? And it   was like more complicated. Some could do it, but  it was more—the barrier of entry was high. Now  

it's very different because of foundation  models and the implications that that has—  Malcolm: With the moment  where the technology is being—  Dario: Democratized. Being democratized. Frankly,  it works better, for classes of problems,   like programming and other things. It’s  really incredibly impressive what it can do.  So the accuracy and the performance of it is much  better. Yeah. And the ease of use and the number  

of use cases we can pursue is much bigger.  So that democratization is a big difference.  Malcolm: You say, when you make an analogy  to the first browsers—if you, if we—to do   another one of these time-travel questions,  back at the beginning of the first browsers, it's safe to say, many of the potential  uses of the internet and such—we hadn't   even begun, we couldn't even anticipate. Dario: Right. Right. Malcolm: Exactly. So we're at the point where the future direction is largely unpredictable. Dario: Yes. Yeah, I think that is right, because it's such a horizontal technology that—  the intersection of the horizontal capability,   which is about expanding our productivity  on tasks that we wouldn't be able to do   efficiently without it—it has to marry, the  use cases that reflect the diversity of human   experience and institutional diversity. So as more and more institutions said,  

you know, I'm focused on agriculture,  you know, to be able to improve seeds,   in these kinds of environments, they'll find  their own context in which—that—matters that   the creators of AI did not anticipate at  the beginning. So I think that that is,   then—the fruit of surprises will be like, why, we  didn't even think that it could be used for that.  And also, clever people will create new  business models associated with that. Like,   it happened with the internet, of course, as well,  and that will be its own source of transformation   and change in its own right. So I think all  of that is yet to unfold, right? What we're  

seeing is this catalyst moment of technology that  works well enough, and it can be democratized.  Malcolm: Yeah. The next sort of conceptual  question: you know, we could loosely understand   or categorize innovations, in terms of their  impact on the kind of, balance of power between   haves and have-nots. Mm-hmm? Some innovations,  you know, obviously, uh, favor those who already   have a—make the rich richer. Some—the—some,  it's a rising tide that lifts all boats,   and some are biased in the other direction. They close the gap between. Is it possible  

to say, to predict, which of those  three categories AI might fall into?  Dario: It's a great question. A first, observation  I would make on your first two categories is that   it will be—both likely be true that the use of AI  will be highly democratized, meaning the number   of people that have access to its power to make  improvements in terms of efficiency and so on   will be fairly universal, and that the ones who  are able to create AI, may be quite concentrated. So if you look at it from the lens of who  creates wealth and value over sustained   periods of time—particularly, say, in a  context like business—I think just being   a user of AI technology is an insufficient  strategy. And the reason for that is, like,  

yes, you will get the immediate productivity  boost of, like, just making API calls and,   that will be a new baseline for everybody.  But you're not accruing value in terms of   representing your data inside the AI in a  way that gives you a sustainable competitive   advantage. So what I always try to tell people  is, don't just be an AI user; be an AI value   creator. And I think that that will have a lot of  consequences in terms of the haves and have-nots,   as an example, and that will apply both to  institutions and regions and countries, etc.  So I think it would be kind  of a mistake, right, to just   develop strategies that are just about usage. Malcolm: Yeah. But to come back to that question  

for a moment, to give you a specific— suppose  I'm a, I'm an industrial farmer in Iowa with 10   million in equipment, and blah, blah, blah. And  I'm comparing it to a subsistence farmer, someone   in the developing world, who's got a cell phone, right. Over the next five years, whose,   whose well-being rises by a greater amount? Dario: Yeah, I think, it's a good question,   but it might be hard to do a one-to-one sort of  like attribution to just one variable in this   case, which is AI. But again, provided  that you have access to a phone, right,   and some way to be able to be connected. I do think—so for example, in that context,   we've developed, we've done work with NASA,  as an example, to build geospatial models,   using some of these new techniques. And  I think, for example, our ability to do   flood prediction—I'll tell you an advantage of why  we'll be a democratization force in that context. 

Before, to build a flood model based on  satellite imagery was actually so onerous   and so complicated and difficult that you would  just target to very specific regions. And then,   obviously, countries prioritize their  own, right? But what we've demonstrated   is actually you can extend that technique  to have like global coverage around that.  So in that context, I would say it's  a force towards democratization—that   everybody sort of would have access  if you have some kind of connectivity. Malcolm: That Iowa farmer might have a flood  model. The guy in the developing world definitely   didn't, and now he's got a shot at getting one. Dario: Yeah, but now he has a shot at getting one. 

So there's aspects of it that—so long as we  provide connectivity and access to it—that   there can be democratization forces. But I'll  give you another example that, that can be quite   concerning, which is language, right? So there's  so much language, in English. And there is sort   of like this reinforcement loop that happens,  that the more you concentrate—because it has   obvious benefits for global communication  and standardization—the more you can enrich   like base AI models based on that capability. If you have very resource-scarce languages,   you tend to develop less powerful AI  with those languages, and so on. So one   has to actually worry and, and focus on the  ability to actually represent, in that case,   language is a piece of culture also in the AI  such that everybody can benefit from it too.  So there's a lot of considerations  in terms of equity about the data,   the data sets that we accrue, and what problems  are we trying to solve. I mean, you mentioned  

agriculture or healthcare and so on. If we only  solve problems that are related to marketing,   as an example, that would be a less rich world in  terms of opportunity than if we incorporate many,   many other broader sets of problems. Malcolm: Yeah. Who do you think—what do   you think are the biggest impediments to the  adoption of, of AI as you would like—as you   think AI ought to be adopted? I mean, if you would  look, what are the sticking points that you would—  Dario: Look, in the end, I'm going to  give a nontechnological answer. The   first one has to do with workflow, right? So even if the technology is very capable,   the organizational change inside a company, to  incorporate into the natural workflow of people on   how we work, is—it's a lesson we have learned over  the last decade is hugely important. Mm-hmm? So   there's a lot of design considerations. There's  a lot of, how do people want to work, right? 

How do they work today? And what is the  natural entry point for AI? So that's   like number one. And then the second  one is, you know—for the broad, uh,   value-creation aspect of it—is the understanding  inside the companies of how you have to curate and create data, to combine it with external  data such that you can have powerful   AI models that actually fit your needs. And that aspect of what it takes to actually   create and curate the data for this modern AI—um,  it's still a work in progress, right? I think part   of the problem that happens very often when I talk  to institutions is that they say, “AI, yeah, yeah,   yeah, I'm doing it, I've been doing it for a long  time.” And the reality is that that answer can   sometimes be a little bit of a cop-out, right? I know you were doing machine learning. You were  

doing some of these things, but actually the  latest version of AI, or what's happened with   foundation models—not only is it very new, it's  very hard to do. And honestly, if you haven't   been, assembling very large teams and spending  hundreds of millions of dollars of compute—in sum,   you're probably not doing it right. You're doing  something else that is in the broad category. And   I think the lessons about what it means to make  this transition to this new wave is still in early   phases of understanding. Malcolm: So what would you say? I want to give you a couple of   examples of people in real-world  positions of responsibility.  Imagine I'm sitting right here. So imagine  that I am the President of a small liberal   arts college. And I come to you and I say,  Darío, I keep hearing about AI. My college  

has— I'm making this much money. If—that  every year, my enrollment's declining,   I feel like this maybe is an opportunity. What  is the opportunity for me? What would you say?  Dario: So it's probably in a couple of segments  around that, right? All one has to do is, well,   what is the implications of this technology inside  the institution itself, inside of the college,   and how we operate? And, can we improve, for  example, efficiency? Like if you're having   very low levels of, of sort of margin to be  able to reinvest, is, you know, you run IT,   you run, infrastructure, you run many things  inside the college. What are the opportunities   to increase the productivity or automate and drive  savings such that you can reinvest that money into   the mission of education, right?—as an example. Malcolm: So number one is operational efficiency.  Dario: Operational efficiency, is a big one. I  think the second one is: within the context of   the college, there's implications for the  educational mission in its own right. How

will—how does a curriculum need to  evolve, or not? What are acceptable   use policies for some of these AI? I don't think—we've all read a lot   about like what can happen in terms of exams  and, and so on, and cheating and not cheating,   or what—are they actually positive elements  of it in terms of how curriculum should be   developed? And professions? Sustain around  that. And then there's another, third,   dimension which is the outward-oriented element  of it, which is like prospective students, right?  So, which is, frankly speaking, a big  use case that is happening right now,   which in the broader industry is called “customer  care” or “client care” or “citizen care.” So—and   this question will be— education. Like, you  know, “Hey, are you reaching the right students?”   Around that—that may apply to the college. How can you create for them, for example,  

an environment to interact with the college, and  answering questions? That could be a chatbot,   or something like that, to learn about it. And  personalization. So I would say there's, like,   at least three lenses with which  I would give advice, right? The—  Malcolm: The second, let's pause on the second  one though, because it's really interesting.  So I really can't assign an essay anymore, can I? Dario: Can I assign an essay? Malcolm: Can I say, “Write me a research paper and come back to me in three weeks?” Can I do that anymore?  Dario: I think you can. Malcolm: How do I do that?

Dario: I think you can. Look, this—so there's two questions around that. I think that if one goes and explains in the context,   like, “What is it? Why are we here? Why are we  in this class? What is the purpose of this?” And,   one starts with assuming, like an element of,  like, decency in people, or people are there,   like, to learn, and so on, and you just give  a disclaimer: “Look, I know that one option   you have is, like, just, put the essay question  and click ‘Go,’ and, like, and give an answer,   you know? But that is not why we’re here, and that  is not the intent of what we’re trying to do.”  So first I would start with the—sort of  like the norms of intent and decency,   and appeal to those, as step number one. Then we  all know that there will be a distribution of use cases—that people like that will come in  one ear and come out of the other and do   that. And,—so for a subset of that, I think the  technology is going to evolve in such a way that,   we will have more and more of the ability to  discern—right?—you know when that has been   AI generated, right? And, created. It won't be  perfect, right? But there's some elements that   you can—imagine inputting the essay, and you  say, “Hey, this is like—it— .” And for example,  

one way you can do that, just to give you  an intuition, you could just have an essay,   uh, that you write with pencil and paper at the  beginning. You get a baseline of what your writing   is like. And then later, when you, generate it,  there'll be obvious differences around what kind   of writing has been generated from the other. Malcolm: Yeah, but you've turned—it's—everything  

you're describing makes sense, but it greatly—in  this, in this respect, at least, it seems to   greatly complicate the life of the teacher.  Whereas the other two use cases seem to kind of   clarify and simplify the role, right? Suddenly,  reaching students, prospective students, sounds   like I can do that much more kind of efficiently. Yeah, I can bring down administration costs,   but the teaching thing is tricky. Dario: Well, until we develop the new norms,   right? I know it's an abused analogy, but  calculators—we deal, we dealt with that too,   right? And, it says, “Well—calculator. What  is the purpose of math? How are we going to  

do this?” and so on. And we have—  Malcolm: Can I tell you my dad's calculator story? Dario: Yes, please.  Malcolm: My father was a mathematician.  Taught mathematics at the University of   Waterloo in Canada. And in the ’70s, when  people started to get pocket calculators,   his students demanded that they be able  to use them. And he said no, and he—they  

took him to the administration and he lost. So he then changed. Completely threw out all   of his old exams. Introduced new exams, where  there was no calculation. It was all like,   “deep think,” you know. Figure out the problem  on a conceptual level and describe it to me. And   they were all—students deeply unhappy that  he had made their lives more complicated. Dario: But it's to your point.  That's the point. To your— 

Malcolm: Point. Right. The result was probably  a better education. Right. He just removed the   element that they could gain with their pocket  calculators. I suppose it's a version of that.  Dario: I think it's a version of that.  And so I think they will develop the   equivalent of what your father did. And I think people say, you know what,   it's like—these kinds of things, everybody's doing  it generically and none of it has any meaning   because all you're doing is pressing buttons. And  like the intent of this was something which was to   teach you how to write or to think or something.  There may be a variant of how we do all of this. 

I mean, obviously some version of that that  has happened is like, okay, we're all going   to sit down and do it with pencil and paper and  no computers in the classroom, but there'll be   other variants of creativity that people will  put forth to say, you know what? You know,   that's a way to solve that problem too. Malcolm: But this is interesting,   because—to stay on this analogy—we're  really talking about a profound rethinking,   just—using a college as an example. A real  profound rethinking of the way—there's no   part of this college that's unaffected by AI, (a).  (B), in one case, I've made everyone's job easier;   in one case I've made—I'm asking us to really  rethink from the ground up what “teaching” means.  In another case, I've automated systems  that I didn't think of. I mean, it's like,  

that's right. That's all—it's not all—that's  a lot to ask someone who got a PhD in medieval   language and literature, 40 years ago. Dario: Yeah, but you know, I'll tell   you a positive sort of development that I'm  seeing. The sciences around this, which is,   you're seeing—as you see more and more examples  of applying AI technology within the context   of like historians too as an example, right? When you have archival and—you know, and you have   all these books, and being able to actually help  you as an assistant, right, around that. But not   only with text now, but with diagrams, right? And,  uh, I've seen it in anthropology too, right? And,   uh, in archaeology, with examples of engravings  and translations and things. That can happen.

So, as you see in diverse fields, people  applying these techniques to advance on how   to do physics or how to do chemistry. They  inspire each other, right? And they say,   how does it apply, to my area? So once, as that  happens, it becomes less of a chore of like,   my God, how do I have to deal with this? But actually, it's triggered by curiosity.   It's triggered by—you know, there'll be  like, you know, faculty that'll be like,   you know what, you know, “Let me explore  what this means for my area.” And they will   adapt it to the local context—to the local,  you know, uh, language, and the profession   itself. So I see that as a positive vector. That is not all going to feel like homework,   you know? It's not going to feel like, “Oh my  God, this is so overwhelming,” but rather to   be very practical, to see what works. What  have I seen others do that is inspiring?   And what am I inspired to do? You know, what—  what is—how is this going to help my career?  I think that that's going to be an interesting  question for, for, you know, those faculty   members, for the students and professionals. Malcolm: Sorry, I'm gonna stick with this example  

alone, because it's really interesting. I'm  curious—following up on what you just said—that   one of the most persistent critiques of academia,  but also of many, of many corporate institutions,   um, in recent years has been “siloing,” right? Different parts of the, of the organization going   off on their own and not speaking to each other  is a potent—is, is a real potential benefit to AI:   the kind of breaking down—a simple  tool for breaking down those kinds   of barriers. Is that a very, is that an  elegant way of sort of summing that up?  Dario: I really think—and I was actually  just having a conversation with a provost   very much on this topic very recently, exactly  on that, which is: all these, this appetite,   right? To collaborate across disciplines. There's a lot of, attempts towards our goal,   right? Creating interdisciplinary centers,  creating dual-degree programs or dual-appointment   programs. But actually, in—a lot of progress in  academia, happens by methodology too. Right? Like  

a new, when some methodology gets adopted—I mean,  the most famous example of that is the scientific method, as an example of that—but when  you have a methodology that gets adopted,   it also provides a way to speak to your  colleagues across different disciplines.  And I think what's happening in AI is, is linked  to that. That within the context of the scientific   method, as an example, the methodology about which  we, about which we do discovery—the role of data,   the role of these neural networks, of how we  actually find proximity of concepts to one   another—it's actually fundamentally different  than how we've traditionally applied it.  So, as we see across more professions, people  applying this methodology is also going to give   some element of common language to each other,  right? And in fact, in this very high-dimensional   representation of information that is present in  neural networks, we may find amazing adjacencies   or connections of themes and topics in ways that  the individual practitioners cannot describe, but   yet will be latent in these large neural networks. We are going to suffer a little bit from  

causality—from the problem of like, “Hey,  what's the root cause of that?” Because I   think one of the unsatisfying aspects that  this methodology will provide is they may   give you answers for which they don't give you  good reasons for where the answers came from.  And then there will be the traditional process  of discovery, of saying, if that is the answer,   what are the reasons? So we're gonna  have to do this sort of hybrid, uh,   way of understanding the world. But I do think  that common layer of AI is a powerful new thing.  Malcolm: Yeah. A couple of random  questions that come to mind as you talk.  In the, in the writer's strike that just ended in  Hollywood, one of the sticking points was how the   studios and writers would treat AI-generated  content—would writers get credit if their   material was somehow the source for AI? But more  broadly, did the writers need protections against   the use of—. I could go on. You know what?  You probably were familiar with all of this.  Had you been—I don't know whether you were,  but had either side called you in for advice   during that? The writers, had the writers  called you and said “Dario, what should we do about AI? And how should we—that  should be reflected in our contract   negotiations?” What would you have told them? Dario: I—the way I think about that is that   I divide it—I would divide it into two parts.  Pieces. First is: what's technically possible,   right? And anticipate scenarios, like, what  can you do with voice cloning? For example,   it is possible there's been, um, dubbing,  right? Like—let's just take that topic,   right? Around the world, there was all these,  folks that would dub people in other languages. 

Well, now you can do these incredible renderings;  I mean, I don't know if you've seen them, where,   you know, you match the lips—it's your original  voice, but speaking any language that you want.   That's the thing. So basically that has a  set of implications around that. I mean,   just to give an example. So I would say: create  a taxonomy, that describes technical capabilities   that we know of today and, uh, applications to the  industry and to examples of like, “Hey, I could   film you for five minutes and I could generate  two hours of content of you and I don't have to,   you know—then if you'll get paid by the hour,  obviously I'm not paying you for the other thing.” 

So I would say “technological capability,” and  then map with their expertise consequences of   how it changes the way they work, or the way  they interact, or the way they negotiate,   and so on. So that would be one element of it. And then the other one is like a   non-technology-related matter, which is an element  of—almost of distributive justice. It's like,   who deserves what? Right? And who has the power  to get what? And, and then that's a completely   different discussion. That is to say, well,  if this is the scenario of what's possible,   you know, what do we want? And what are we able  to get? And, I think that that's a different   discussion, which is, as old as life. Malcolm: Which one do you do first? I think it is very helpful to have an understanding of what's possible and how it  changes the landscape, uh, as part of a broader,   uh, discussion—right?— and a broader negotiation.  Because, you also have to see the opportunities,  

because there will be a lot of ground to  say, “If we can do it in this way, and we   can all be that much more efficient in getting  this piece worked on or this filming done....” But we have a reasonable agreement about how  we—both sides—benefit from it, right? Then   that's a win-win for everybody, right? So that's  a—I think that would be a golden triangle, right?  Malcolm: Here's my reading, and I would  like you to correct me if I'm wrong. And   I'm likely to be wrong. Uh, when I looked at  that strike, I said, “If they're worried about  

AI—the writers are worried about AI. That seems  silly. It should be the studios who are worried   about the economic impact of AI.” Doesn't,  in the long run—AI puts the studios out of   business long before it puts the writers out  of business. I only need the studio because  

the costs of production are as high as the sky  and the costs of production are overwhelming.  And—whereas if I don't, if I have a  tool which brings, introduces massive   technological efficiencies to the production  of movies, then I don't need—why do I need a   studio? Why would they be the scared ones? Dario: Or maybe—or maybe you need a, like,   a different kind of studio. Or a different  kind of studio. A different kind of studio.  Malcolm: What do you mean? In this strike,  the frightened ones were the writers and   were the studios. Wasn't that backwards? Dario: I haven't thought about it. But the   implications of it—it goes back to what we  were talking about before. The implications,   because they're so horizontal—it  is right to think about it. Like,   what does it do to the studios as well, right? But then, the reason why that happens is that   it's the order of either negotiations or who  first got concerned about it and did something   about it—right?— which is in the context of  the strike. Um, you know, I don't know what the  

equivalent conversations are going on inside  the studio and whether they have a war room   saying what this is going to mean to us, right? But it doesn't get exercised through a strike,   but maybe through a task force inside, the  companies, about what are they going to do, right? Malcolm: Well—and to go back to your  thing, you said the first thing you do   is you make a list of what technological  capabilities are, but don't technological   capabilities change every—? I mean, they do. You're racing ahead so fast. So you can't—can   you have a contract? Again, I'm sorry  for getting into a little weeds here,   but this is interesting. Can you have a—you can't  have a five-year contract if the contract is based   on an assessment of technological capabilities  in 2023. Because by the time we get to 2028,   2028, it's totally different, right? Dario: Yeah, but like, where I was going   is like—there are some, abstractions around  that. It’s like, what can we do with my image,  

right? Like, if I generally get the category,  that my image can be reproduced, generated,   contents, and so on, it’s like, let’s talk about  the abstract notion about who has rights to that,   or do we both get to benefit from that? If you get that straight, yes, the nature   of how the image gets altered, created, or  something—it will change underneath, but the   concept will stay the same. And, uh, so I think  what’s important is to get the categories right.  Malcolm: Yeah. Yeah. If you had to—if you had to  think about the biggest technological obstacle,   revolutions of the postwar era—last 75 years—we  can all come up with a list. Actually, it’s really   fun to come up with a list. I was thinking about  this when we were, you know—containerized shipping   is my favorite. The green revolution.  The internet. Where is AI in that list?  Dario: So I would put it first. In that context that you put forth,  

since World War II, undoubtedly, like, computing  as a category is one of those trajectories that   has reshaped, right, our world. And I  think within computing, I would say,   the role that semiconductors have had has been  incredibly defining. I would say AI is the second   example of that as a core architecture, uh, that  is going to have an equivalent level of impact.  And then the third leg I would put to  that equation will be quantum. Quantum  

information. And that’s sort of like—I like  to summarize that the future of computing is bits, neurons, and qubits. And it’s that  idea of high-precision computation—the   world of neural networks and artificial  intelligence and the world of quantum.  And the combination of those things  is going to be the defining force   of the next 100 years in that category of  computing. But it makes the list for sure. 

Malcolm: If it’s that high up on the list, this  is a total hypothetical. Would you—if you were   starting over; if you were starting at IBM right  now—would you say, “Oh, our AI act operations   actually should be way bigger”? Like, how many thousands   of people working for you? Dario: So within the research division,   uh, it’s about like 3,500 scientists. Malcolm: So in a perfect world, would you,   if it’s that big, isn’t that too small a group? Dario: Yeah. Well, that’s like in the research   division. I mean, IBM overall, there’s tens  of thousands of people working on that.  Malcolm: We’re talking, we’re talking about—but  I mean, like, so, starting from—first,   so you have a—you’ve, we’ve got a technology  that you’re ranking with compute and, you know,   up there with it in terms of a world changer.  Are we—so what I'm basically asking is,  

are we underinvested in this future? Dario: No, but so—so yeah,   it’s a, it’s a good question. So like what I would say is that   I think we should segment. How many people do you  need on the creation of the technology itself? And   what is the right size of research and engineers  and compute to do that? And how many people do you   need in the sort of application of the technology  to create better products, to deliver services   and consulting, and then ultimately to diffuse it  through, you know, sort of all spheres of society?  And the numbers are very different, and that  is not different than anywhere else. I mean,   I mean, if you give examples of—since you were  talking about, in the context of World War II,   how many people does it take to create, an atomic  weapon as an example. It’s a large number. I mean,   it wasn’t just Los Alamos. There’s a lot  of people in Oakland. It’s a large number,  

but it wasn’t a million  people, right? Um, so, so you could have highly concentrated teams of people  that with enough resources can do extraordinary   scientific and technological achievements. And  that’s always—by definition, is going to be, uh,   1 percent compared to the total volume that  it’s going to require to then deal with it.  Malcolm: Yeah. But the application side is infinite, almost. Dario: That’s exactly—so that is where, like,   in the end, the bottleneck really is. So, with,  you know, thousands of scientists and engineers,   you can create world-class AI. Right? And, you  don’t need 10,000 to be able to create the large  

language model and the generative model and so on. But you need thousands, and you need,   you know, a very significant amount of compute  and data. You need that. The rest is, “Okay, I,   build software,” “I build databases,” or “I build  a software product that allows you to do inventory   management,” or “I build, a photo editor,” and  so on. Now that product, incorporating the AI,   modifying, expanding it, and so on—well, now  you’re talking about the entire software industry.  So now you're talking about millions of people,  right, who are required to bring AI into their   products. Then you go a step beyond the technology  creators in terms of software and you say, well,  

okay, now what? The skills to help organizations  go and deploy it in the Department of,   you know, the Interior, right? And then I said, okay, well,   now you need like consultants and experts  and people to work there to integrate into   the workflow. So now you’re talking into the many  tens of millions of people around that. So I see   it as these concentric circles of it. But to some  degree in many of these core technology areas,   just saying like, well, I need a team of like a  hundred thousand people to create, like, AI, or a,   or a new transistor or a new quantum computer— It’s actually a diminishing return,   right? In the end, like, too many people  connecting with each other is very difficult.  Malcolm: But, on the application side,  it was just—think of our example of that   college. Just the task of sitting down  with a faculty and working with them   to reimagine what they do with this,  with this new set of tools in mind,   with the understanding that the students  coming in are probably going to know   more about it than they do—that alone—I mean,  that’s a, that is a Herculean people problem.

Dario: It’s a people problem. Yeah,  that’s why I started in terms of the   barriers of adoption of that. I mean, in the context of   IBM, an, an example—that's why we have a  consulting organization, IBM Consulting,   that complements IBM Technology, and the  IBM Consulting organization has over 150,000   employees. Because of this question, right?  Because you have to sit down and you say,   okay, what problem are you trying to solve?  What is the methodology we're going to use?   And here's the technology options that we have  to be able to bring to the table. In the end,   the adoption across, uh, our society will be  limited by this part. The technology is going  

to make it easier, more cost- effective  to implement those, uh, solutions. But   you first have to think about what you want to  do, how you're going to do it, and how you're   going to bring it into the life of this—in this  context, a faculty member, or, uh, you know, the   administrator and so on in this college, right? Malcolm: With that Hollywood, that, that notion,   I thought, which was absolutely, I thought really  interesting that, in a Hollywood strike, you have   to have this conversation about—a distributive  justice conversation about how do we—that's,   it's a really hard conversation, right, to  have in a—so this brings me to my next point,   which is that you—we were talking backstage. You have two daughters, one in college,   one about to go to college. Darío: That's right. Malcolm: So, they're both science minded. 

Darío: Yeah. Malcolm: So tell me about   the conversations you have with your daughters.  You have a unique conversation with your daughters   because your conversa—your advice to them is,  is influenced by what you do for a living.  Darío: Yes, it's true. Malcolm: Did you warn your   daughters away from certain fields? Did you  say, “Whatever you do, don't be”—you know? Dario: No, no, no, no. That's not my style.  I mean, for me, no. I try not to be like,   preachy about that. So for me it was just  about showing by example of things I love,  

right? And, things I care about. And then, bringing them to the lab   and seeing things, and then the natural  conversations of things I'm working on,   or interesting people I meet. So, to the extent  that they have chosen that—and obviously this   has an influence on them—it has been through  seeing it, perhaps through my eyes, right?  And what you see me do, and that  I like my profession. Right? 

Malcolm: But one of your daughters, you said,  is thinking that she wants to be a doctor. But   being a doctor in a post-AI world is surely a  very different proposition than being a doctor   in a pre-AI world. Do you think—have you  tried to prepare her for that difference?   Have you explained to her what you think will  happen to this profession she might enter?  Dario: Yeah. I mean, not in like, you  know, incredible amount of detail, but yes,   at the level of understanding what is changing,  like this lens of the—information lens with which   you can look at the world and what is possible,  uh, and what it can do, like what is our role and   what is the role of the technology and how that  shapes at that level of abstraction, for sure.  But not at the level of like,  don't be a radiologist, you know,   because this is what we want for you. Malcolm: I was going to say, if you,  

if you're unhappy with your current job, you could  do a podcast called Parenting Tips with Darío,   which is just, “an AI person gives you advice on  what your kids should do based on exactly this.”  Like, “Should I be a radiologist? Darío,  tell me.” Like, it seems to be a really   important question. Darío: Yeah.  Malcolm:, Let me ask this question in a  more—I'm joking, but in a more serious way,   surely it would—if—I don't mean to  use your daughter as an example,   but let's imagine we're giving advice to somebody  who wants to enter medicine. A really useful

conversation to have is, what are the skills  that are—will be—most prized in that profession,   yeah, fifteen years from now, and are they  different from the skills that are prized now?  How would you answer that question? Darío: Yeah, I think, for example—this is,   goes back to how is the scientific method on,  in this context, like the practice of medicine,   going to change? I think we will see more changes  in how we practice the scientific method and   so on as a consequence of what is happening  with the world of computing and information,   how we represent information, how we  represent knowledge, how we extract   meaning from knowledge as a method, uh,  than we have seen in the last 200 years.  So therefore, what I would like strongly  to encourage is not about, like, hey,   use this tool for doing this or doing that, but  in the curriculum itself, in understanding how   we do problem solving in the age of like  data and data representation and so on;   that needs to be embedded in the curriculum of  everybody. You know, that is, I would say actually   quite horizontally, but certainly in the context  of medicine and scientists and so on, for sure. 

And to the extent that that gets ingrained, that  will give us a lens that no matter what specialty   they go with in medicine, they will say, actually,  the way I want to be able to tackle improving the   quality of care, the way to do that is—in addition  to all the elements that we have practiced in our,   in the field of medicine—is this new lens. And  are we representing the data the right way? Do   we have the right tools to be able to represent  that knowledge? Am I incorporating that in my   own—sort of with my own knowledge in a way that  gives me better outcomes, right? Do I have the   rigor of benchmarking to, and quality of, the  results? So that is what needs to be incorporated.  Malcolm: How, in a perfect world, if I asked  you to, your team to rewrite the curriculum   for American medical schools, how dramatic  a revision is that? Are we tinkering with   10 percent of the curriculum or are  we tinkering with 50 percent of it?  Darío: I think there would be, a subset of classes  that is about the method—the methodology. What has   changed. Like, like, have this lens of it  to understand. And then within each class,  

that methodology will represent  something that is embedded in it, right? So it will be substantive, but it  doesn't mean replacing the specialization   and the context and the knowledge of each domain. But I do think everybody should have sort of a   basic knowledge of the horizontal, right? What  is it? How does it work? What tools you have,   what is the technology, and like, you know,  what are the dos and don'ts around that. And   in every area, you say—“That thing that you  learn? This is how it applies to, uh, anatomy,   and this is how it applies to radiology,” if  you're studying that. “Or, this is how you apply,   in the context of discovery—right?— of cell  structure. And this is how we can use it.”   Or “protein folding, and this is how it, it  does—.” So that way, you'll see a connecting  

tissue through, uh, throughout the whole thing. Malcolm: Yeah. I mean, I would add to that. It's   also this incredible opportunity to do what  doctors are supposed to do but don't have time   to do now, which is, they're so consumed with  figuring out what's wrong with you that they   have little time to talk about the implications  of the diagnosis. And what we really want are—if   we can free them of some of the burden of what is  actually quite a prosaic question of “What's wrong   with you?” and leave the hard human thing of let  me—should you be scared or hopeful? Should you—,   what do you need to do? What—let me put this in  the context of all the patients I've seen. That   conversation, which is the most important one, is  the one that's—seems to me. So like if I had to,  

I would add, if we were reimagining the curriculum  of med school, I'd like—with whatever—by the way,   very little time. Maybe we have to  add two more years to med school.  But like a whole—that's not gonna be popular.  But the whole thing about bringing back the   human side of, yeah, you know, now  if I can give you ten more minutes,   how do you use that ten more minutes? Darío: But in that, in that   reconceptualization that you just did  is what we should be doing around that.  Because I think the debate as to like, “Well, am  I gonna need doctors or not?” is actually a not   very useful debate. But rather this other question  is “How is your time being spent? What problems  

are you getting stuck?” I mean, I generalize  this by like the obvious observation that if   you look around in our professions, in our daily  lives, we have not run out of problems to solve. So as—an example of that is, hey, if I'm spending  all my time trying to do diagnosis, and I could do   that ten times faster, and it allowed me actually  to go and, um, you know, and take care of the   patients and all The next steps and what we have  to do about it. That's probably a trade-off that a   lot of doctors would take—would take, right? Yeah. And then you say, well, to what degree does it   allow me to do that? And I can do these other  things and these other things that are critically   important for my profession around that. So  when you actually become less abstract, and   like we get past the futile conversation of like,  “Oh, there's no more jobs and AI's gonna take it,   all of it,” which is kind of nonsense, is: you go  back to say, in practice, in your context, right,   for you, what does it mean? How do you work? What  can you do differently around that? Actually,   that's a much richer conversation. And very often  we would find ourselves—that there's a portion of  

the work we do that we say, “I would rather do  less of that. This, this other part I, I like a   lot. And if it is possible that technology could  help us make that trade- off, I'll take it in a   heartbeat.” Now, poorly implemented technology can  also create another problem. You say, hey, this   was supposed to solve things, but the way it's  being implemented is not helping me, right? It's   making my life much more miserable, or so on, or  I've lost connection in how I used to work, etc.  So that is why design is so important. That is  why also workflow is so important in being able to   solve these problems. But it begins by, you know,  going from the intergalactic to the reality of it,  

of that faculty member in the liberal arts  college or a practitioner in medicine in a   hospital and what it means for them, right? Malcolm: Mm-hmm. Yeah. What struck me, Darío,   throughout our conversation is, um, how much of  this revolution is nontechnical. ’Cause to say,   “You guys are doing the technical thing  here, but the real, the revolution is   going to require a whole range of people doing  things that have nothing to do with software,   that have to do with working out new, new  human arrangements”—talking about that,   I mean, I keep coming back to the Hollywood strike  thing: that you have to have a conversation about   our values as creators of movies; how are  we going to divide up the credit, and the—  Dario: Exactly right! Malcolm: Like that’s a conversation about philosophy, and, Darío: it's in the grand tradition of why,   a liberal education is so important in  the, the broadest possible sense, right?  There's no common conception of the good,  right? That is always a contested, uh,   dialogue that happens within our society. And  technology is going to fit in that context too, right? So that's why personally, as a  philosophy, I'm not a technological determiner.  

Right? And I don't like when colleagues in my  profession, right, start saying like, well, this   is the way the technology is going to be, and by  consequence, this is how society is going to be.  I'm like, that's a highly contested goal, and  if you want to enter into a realm of politics   or the realm of other ones, go and stand  up on a stool and discuss whether that's   what society wants. You will find that it's a  huge diversity of, of opinions and perspective,   and that's what makes, you know, uh, you know,  in a democracy, the richness of our society.  And in the end, that is going to be the  centerpiece of the conversation. What do we  

want? You know, who gets what? And so on, and  that is—actually, I don't think it's anything   negative. That's as it should be. Because in  the end, it's anchor of who we want as humans,   as friends, family, citizens, and we have many  overlapping sets of responsibilities, right?  And as a technology creator, my  only responsibility is not just as   a scientist and a technology creator; I'm  also a member of a family, I'm a citizen,   and I have many other things that I care  about. And I think that—that sometimes   in the debate of the technological determinists,  they start now butting into what is the realm of   justice and society and philosophy and democracy. And that's where they get the most uncomfortable,   because it's like—I'm just telling you  like, you know, uh, what's possible. And  

when there's pushback, it's like, yeah,  but, but now we're talking about how we   live. And how we work and how much I get paid  or not paid. So that technology is important.  Technology shapes our conversation. But  we're gonna have the conversation with a   different language. As it should be. And  technologies need to get accustomed to—if   they want to participate in that world with the  broad consequences, hey, get accustomed to deal   with the complexity of that world. Of politics,  society, institutions, unions, all that stuff. 

And, you know, you can't be, like, whiny  about it. It's like, “They're not adopting   my technology.” That's what it takes  to bring technology into the world. Malcolm: Yeah, well said. Thank you, Darío, for  this wonderful conversation. Thank you, to all of   you for coming and listening. And, thank you. Darío: Thank you. Malcolm: Darío Gil transformed how  I think about the future of AI.  He explained to me how huge of  a leap it was when we went from  chess-playing models to language-learning models. And he talked about how we still  

have a lot of room to grow. That’s why it’s important that we get things right.  The future of AI is impossible to predict.  But the technology has so much potential   in every industry. Zooming into an academic  or a medical setting showed just how close  

we are to the widespread adoption of AI. Even  Hollywood is being forced to figure this out.  Institutions of all sorts will have to be at the  forefront of integration in order to unlock the   full power of AI thoughtfully and responsibly.  Humans have the power and the responsibility to   shape the tech for our world. I, for one,  am excited to see how things play out. Smart Talks with IBM is produced  by Matt Romano, Joey Fischground,   David Zha, and Jacob Goldstein.  We’re edited by Lidia Jean Kott.

Our engineers are Jason Gambrell,  Sarah Bruguiere, and   Ben Tolliday. Theme song by Gramoscope. Special thanks to Andy Kelly, Kathy Callaghan,   and the EightBar and IBM teams, as  well as the Pushkin marketing team.  Smart Talks with IBM is a production  of Pushkin Industries and Ruby Studio   at iHeartMedia. To find more Pushkin podcasts,  listen on the iHeartRadio app, Apple Podcasts,   or wherever you listen to podcasts. I’m Malcolm Gladwell.  This is a paid advertisement from IBM.

2024-05-30 11:00

Show Video

Other news