Arvind Krishna, Chief Executive Officer, IBM
Cristiano Amon, Chief Executive Officer, Qualcomm
Moderator: Polina Pompliano, Deputy Director, Live Events, FORTUNE; Founder, The Profile


CEO Initiative 2023: Navigating The New Frontiers Of Business And AI

Conferences

I feel like every few years, Ai makes a splash but this time feels a little bit different. Can you share with us more about your organization and how it currently uses Ai Arvin? Happy look. A has been making a splash for 30 years. Every 10 years, it goes up, it goes down. We've had the winter of AI but this time is actually significantly different. You got to go to the underlying science. These technologies of large language models have fundamentally changed the cost economics of AI you can now while it takes a lot of infrastructure to get your base model on each incremental task is an evening or a weekend of work that is fundamentally different than 100 phds and six months of work. That's a really big difference. The multimodality is a different one. And I think that the use cases of which I'm sure we'll touch on. Many are across the fabric of almost everything we do, not just in tech companies but others. One tidbit in some of our hr transactional services mo people raises moving people, employment verification. We could do 90% efficiency using these technologies. So I'll leave you with that Tisa. Yes, look, um I, I agree with uh Ivin. I think A I has been with us for a while but there are a couple of things that really accelerating in the process. Um You know, the I think the A I revolution we see right now is is basically enabled by, by the development in three underlining technologies. I think computing performance, especially as semiconductors are increasing drama dramatically. You have the availability of data and you have broadband both wired as well as wireless. Combining those three things to create an environment that you actually can deploy A I very pervasively. And I I think we are at a moment that uh we're going to see fundamental changes in actually in user behavior, both a consumer and the enterprise. You know, I like to use this example, like I want to say this could be could be as powerful as one day you have your future phone and then you saw this thing called the smartphone. And you said, well, this is great, this is new and I I think that's probably that order of magnitude change when you think about um changing user experience. But also if you don't want to believe in that, it's almost like the underlining technology of GP F. And then all of a sudden you saw something like Uber as a very interesting application that was enabled by it. I think that's how this thing is going to develop and it's going to accelerate very fast to your question about using A I. We've been using A I for a number of years. I think uh just provide us integrated into the integrated in how we develop our products. I'll give you some examples when we develop our five G modem. Now, a significant percentage of development is using A I just processing a lot of data and helping us develop our modem. We do that for image signal processes for a camera and even gen A I we've been utilizing internally. We have developed our our ability to engineers, consulting documentation, remembering documentation. We've been using GA I to do this with precision and I think the applications is going to continue to expand actually as a company, we started uh to be using uh some of the Microsoft co pilot as well. And we have seen significant gains of productivity and avid you've written pieces uh and talked about how A I will affect white collar jobs. First. Can you explain this a bit more and how you claim that it will help workers instead of displacing them? Absolutely. Look if you look at where all of these technologies can be deployed and you look at it through automation. The first thing you can automate is a repetitive white collar job. People often confuse blue collar work. Blue collar work has had the attacks from robotics and automation, but it's white collar work that will get displaced and I'm using the word displaced, but only for the lower level task. That means for the majority of people, it's a productivity augmentation. It's sitting with them. It's not really replacing the whole human because nobody has 100% of that repetitive work. Now for the 10 20% it does. There is also going to be 30% new work that is created the same way Khan, I used the example of smartphone, I'll use the internet in 1995. Nobody thought there'd be 5 million web designers. There are dito there are going to be people who are going to come into new roles with the advent of generative A I and large language models. Second, I think that the the issue is different in the global south versus developed countries in developed countries. We have a shortage of labor, not a, not a, not a abundance of labor. So we need these technologies just in order to keep our quality of life and the GDP growth going on. Then in the other places as you begin to have more and more customer service, customer experience, the number of roles in those jobs is going to increase. Not to even touch on what the last panel was talking about in health care where the number of workers needed is going to only increase as aging goes on in all societies. So I do think that there will be 10 to 20% displacement in some white collar jobs more than made up by the increase. So we talk about our programmers getting 30% more productive. I don't intend to get rid of a single one if I'll get more because that makes the task of making things that are using more software more productive. That means you can make more of them because they can get more of a market. So that is why people mistake productivity with job displacement. It's not job displacement. It actually makes it more attractive to have those roles. Have you made any changes to departments within IBM? Because of this? Absolutely, we have decreased in certain back office functions and we have increased in more value creating activities. Can you give an example? Yes, I began by saying in our human resources function, a few 100 people were no longer needed in the back office. That is a journey that takes 3 to 4 years. It's not a big bang on one day, but over 3 to 4 years that went away the number of people in uh software delivery and in uh client facing sales went up because we could have more attractive offerings in those areas. And by the way, the increase I think was like 8000, the decrease was like 800. So just to be clear, I mean, as you get more productive, generally people invest more. That's the fundamental issue in business. Yeah, I'll come to you guys with questions if you have one, please raise your hand. We'll get a mic there. Um In one minute Cristiano, you have said that in order for generative A I to really scale and get into people's homes, it cannot all happen on the cloud. It has to happen on the device itself. Can you explain that a bit more? Absolutely. Actually, I would like to spend a little bit of time on this because um this uh it's intuitive like when you look at, for example, especially chat C BT um made uh gen A I very popular and everybody understand it and it's intuitive. You think you go to a cloud and you do a search and, and you get results, but the reality is you should think about A I a little bit different A I is going to develop in a particular way in the cloud, but it's going to develop on devices a little bit different as a new form of computing. Think about anything that requires real time latency, you need a response right away or you need to know the context that you inserted or you have the ability uh you want to protect the data with privacy, then you're gonna have A I running pervasively on the devices and I'll give an example. So I'm going to go back and talk about phones, but I can talk about phones, you know P CS and cars, but phones, it's, we used to have when you look at your smartphone today, you have a computing platform, you have an OS you have applications and then you have the human and the human go and touch his different apps and get data from one app to the other. Now, think about this GE I running and understanding everything that you're doing on the phone and running pervasively and try to predict your every action you take. So then all of a sudden you have the ability to have the co pilot for everything you do. For example, you and I could be texting it something very common with of our phones. I can tell you about my vacation somewhere immediately. The phone can bring my gallery app, identify the pictures from the vacation, ask if I want to share with you or we talk about getting together, bring the calendar app, talks about availability. So think about a co pilot of everything and those things are running pervasively. But then you have some interesting consumer applications as well that happens on the device. Like you take a picture of your phone probably multiple times a day. You take a picture. I said I don't want, I don't want Irwin in this picture. I want to delete him. I want to change the background from this room to Venice and, and then you send it to somebody else. And I think you're going to do this by running many models now on a PC. You think about the use of productivity and, and think about the co pilot exercise and you can do a lot of those things running that on the device based on your data. The car is the one that I'm super excited about because natural language communication is perfect user interface when you are behind the wheel, especially when you think about uh navigation something as simple as I'm going from A to B. But there are so many other things that you want to do or you have done or is in your agenda and the gen A I will be able to plan your trip based on that. So the way I see it is we are seeing two parallel developments happening just because you now have the computing power data and you have broadband, you have seen the development of that in the cloud and you've seen that development on device and they have a point that they intersect, they intersect when you can run models from the cloud in the device. If a model is very well trained. And that's one of the things we have been doing, we of show running billion parameter models running on a phone, we're going to multiple tens of billion parameters and phones and P CS and cars. When you have that model well trained, you have the ability to take advantage of the distributing computer that exists in everybody's devices and, and that's a massive amount of computing power and it saves a lot of cost actually on the cloud. So we're seeing that application as well. Sorry for the long answer, but we're very passionate about this, but there's a whole A I on device coming. Awesome. Do we have any questions? Yes. Sorry. How are you changing? How you are recruiting um for people as you look five years down, 10 years down the road. And how do you think the university should change its curriculum to meet this new A I era? I'll jump in first. I don't think that people, there is inventing A I and there is using A I for using A I. The number of people with skills is massive. We just have to give them a little bit of training on how to use it and deploy it. So there, I think people with skills across the social sciences, across the hard sciences, engineering, they applicable. Inventing A I is where you have a poet of skills today. And that is the classic, either hard physics or computer science or math training. That is typically there. But inventing A I, you need thousands of people in the country, not hundreds of thousands, thousands using A I, we need millions. And I think that a lot more attention ought to be on the using as opposed to inventing I think for universities, I will tell you that uh training people on this the same way as software engineering became a discipline 15 years after the fact, maybe universities should jump in hard in terms of training people on how to use A I and also make them aware of what are the risks on A I. That is something which I think a lot more has to be taught on, maybe just to add one thing to that. And uh when we talk about the many different benefits of A I, one thing that if you democratize A I, it has an incredible powerful um ability to accelerate training and actually to distribute knowledge to the point that you can have workers that otherwise will not be able to be qualified to do certain tasks, be able to do it with the copilot with them. So, and I think that's another interesting aspect of this transition. Any other questions? OK. Um You are both kind of very massive global companies and in a more global context. How do you see China positioning itself in the A I landscape? And, and what impact does do China's A I capability have on competition or your company, Irvin? Well, look, um I wish that there was more peace and global trade. So I'll begin with a basic statement there. I kind of agree with what Senator Warner said about. Let's separate the politics from the, from the nation. Now, all that said China has advanced in A I, let's not make any illusions around that. Their ability to use large private data sets, their ability for the government to push forward on certain use cases are probably not use cases that we are going to push forward here. Now, unfortunately, as this more of a tech war there is going to be more and more development that is splintered across the two nations. In terms of these use cases, China is not going to sit still in terms of what's happening on five G and six G as well as on the intersection of AI and quantum and all those those are going to happen. Now, I'm going to sound a little bit contrarian. I actually believe when I look at it right now that at the fundamental advanced level, the US is ahead. How much ahead we can debate? Is it six months ahead or two years ahead? But the fact that chart GPT came out here got copied than six months. The fact that generative AI came out of places like Stanford and MIT and Google and Facebook is a fact that I think we should all take pride in how much longer we can maintain that lead is an interesting question by the one thing I want to jump in on Christian. I mentioned chip design over 3/4 of the work in the chip designs that IBM does is done by A I today using some of the same technologies that Christian's team does. And those chip design tool companies are also based here in the United States largely. Yeah, maybe. Oh I add one thing a I it's a very pervasive technology that is going to be part of many, many industries. So as a result, I think as industries evolve, regardless where they are, uh you're gonna see more and more of a role of A I look um from a technology standpoint, um it just focus on the technology, not the politics aspect, you know, uh technology seems very uh it is difficult for the technology to see boundaries, especially because the world became much smaller and we are in a very global interconnected economy. I think you have industries that have some point of interconnection and you see A I coming in different ways. Now there's different ways to talk about A I, you can have, for example, A I that you probably already have it in your phone when you take a picture at night and all of a sudden the picture looks very bright like in daylight. The A I figuring out this is the sky should be blue, this is a tree should be green, it's already happening. It's very different and for you to have massive ability of doing training on different models. So I think there's different type of applications, I think it will be very difficult, I think to be able to have a blank statement on uh where A I is going to go and how to regulate that. But the most important thing I think in this whole conversation is what Ivin alluded to. Uh you know, the the United States has a significant lead in semiconductors. I think we often talk about the fact that we wish we have more semiconductor manufacturing in the country. We we don't have as much as we need on manufacturing, but on design, the majority of the global leaders on chip designs are American companies and A I. It's going to require a number of industries that the US as a leadership. So the most important thing we can do is to make sure the US remains the number one place to attract talent because there's some finite number of individuals that can help create a is on those technologies that we have the ability to remain competitive and we maintain our innovation engine which creates that leadership position in the first place. Perfect and note to end on. Thank you, gentlemen, so much. Thank you so much.