Ruth Porat - President of Alphabet | Podcast | In Good Company | Norges Bank Investment Management

Who is going to win the AI race? Well, we feel we're in a really strong position for a number of reasons. I think when you look at the history of Google, focused on AI, we started many, many years ago. And at this point, we have a very differentiated approach, which starts with the extraordinary team we have, led by Demis Hassabis, who obviously just won the Nobel Prize. That goes to the strength of the models. You look at what we're doing on chips with our TPUs, and then we're already really using it across our various platforms, so billions of people are benefiting from AI. So this full-stack approach we think is an important element. That being said, what's exciting to see is how much innovation there is broadly, and so what we're looking at is the opportunity. I think collectively globally to have an unlock from the upside given the innovation that we're seeing not just at Google but at other places.

Well, the reason I started with what we're really proud of, our full-stack approach from talent and models to the platforms and chips, is it really is taking this full-stack approach. And I think very importantly, this intense focus on innovation and continuing to push forth. You know, one of the things that really struck me when Demis was awarded the Nobel Prize this last fall, and he was asked what was it that really started him on this journey, and he said what motivates him is to take on the most intractable, what were thought to be the most intractable problems facing humanity. And I think that ethos is another really important element of how one continues to drive forth and have the biggest impact with AI.

I think generative AI is continuing to enable us and others to expand the quality of what is being delivered across a wide set of opportunities. So just to give you two examples, you know, I'm often asked where we are investing extensively, and probably not surprisingly, you know this well given how close you are to all that we've been doing, but search remains so core to who we are. And we've evolved search meaningfully over the years, and we're continuing to apply generative AI to that experience that one has when you search. And what we're finding is that it opens the types of queries that are being explored, it goes for a deeper, richer engagement. So that's one area. The other area that's really important is what we're seeing on the enterprise side, and the ability to help companies and the public sector transform their businesses and their approach, whether it's engaging with customers or on the efficiency side or on risk analytics or with constituents. Each of those provides opportunity for monetization.

Well, a search, Google search for decades has really stood for extraordinary quality. It's what everybody around the globe expects of us. When you go to Google, a quality answer very rapidly surfaced for you in the most potent way. And one of the very important questions for us, Sundar talked about this in the early days of generative AI internally, we were all talking about the risk of hallucination, and that term now is very well known quite broadly. And one of the concerns is if you, in the middle of the night, wake up, your child is sick, you want to figure out how much Tylenol to give to a three-year-old, there can be no margin for error. That's what our brand stands for. And so it's very important to us to make sure that as we were evolving and applying generative AI, we did it in a way that was consistent with the quality that's expected appropriately from Google. And I appreciate your question because what you've seen is the really ongoing momentum in models and introducing models more broadly externally. What we've done with, for example, AI Overviews, where when you search you'll get this kind of an AI cockpit, is the way I like to think about it. You're seeing more and things like a model advancement that can be applied in other applications. And so we're excited about the momentum that people have seen through 2024 and what is ahead in 2025.

So that's one area that I know Sundar and Demis and the team are really excited about, which is how can AI actually be an agent working on our behalf? How can it help you do some of the administrative tasks that are going to make your life easier and actually free you up to do something else? Book a reservation for you, research something for you. How can it even be applied, for example, in science to do some of the basic inquiries? So set it out to do something to work on your behalf.

You know, that's a great question. As it relates broadly to AI, I think one of the most important things for all of us is that AI can really be an assist. It can be operating leverage for every one of us. When I started my career on Wall Street, how was it as you were moving from big computers to laptops to phone? This becomes operating leverage if appropriately applied. And so what one scientist once said to me, that they view this as augmented intelligence, not artificial intelligence, and wish that had been the word that was used. That's what it is for each one of us. And I think what's really important is to understand that in fact it's not that we will be replaced by AI in our roles, but we can be replaced by someone who's using AI, who's getting that operating leverage, if we're not. And so it's so important for people to just start experimenting, just start playing, so that you're on that learning journey.

Well, we work very closely with Nvidia, they're a strong partner of ours, so we use both GPUs and TPUs. And many years ago, the team started on this journey of developing our own chips, TPUs, to address some of the specific requirements that we felt we needed. And what we found is that we continue to have the type of performance that we're looking for, in particular in training. And as we've advanced the TPUs, what we're seeing is we're continuing to drive greater efficiency in all elements of energy requirements with TPUs. So there's a cost benefit for us and an impact benefit. And so we're using both in the fleet.

So open source has been core to Google really since inception. If you think about something like Android, which is such an important operating system globally, it reflects who we are. When you think about the Transformer paper, which has been so critical for anybody who's thinking about how do you build in this AI world, what's the next step to the next iteration. So open source has been important to us. At the same time, this is a really powerful technology, and so what we're looking at is how does one maximize upside but have the appropriate guardrails and controls over certain elements of it to ensure that you're protecting how it's used and where it's used. And getting that balance right will continue to be very important for all of us.

Oh, so when we look at AI, I think there are four primary areas that are really exciting. First is the economic upside. You know, globally the estimates are there can be $20 trillion added to GDP in the next decade, again if appropriately executed, if adopted across industry. When you think about the benefit to society, if in fact we have anything close to that economic uplift, I get excited about it. It's not to be taken for granted, as I said, it requires what economists call diffusion across industry, adoption, a radical rethink of many elements of the way the public sector and the private sector works. Excited about that. Very excited about the breakthroughs in science. What Demis Hassabis and John Jumper awarded the Nobel this past fall, you know that was for something called AlphaFold, which you're familiar with, which has been described as the greatest contribution to drug discovery. That is exciting. Then just even sitting here today, the practical applications of AI in healthcare, in education are really exciting. So I would put those at the top of the list.

So we've been working on quantum AI, quantum computing for quite some time, well over I think a decade at this point. And what's really exciting is the computational capabilities with quantum. So the Willow chip is able to handle a computation in less than 5 minutes that previously on the best supercomputers on the planet today would take 10 septillion years, which even I had to Google, it's 24 zeros on the back of it. So what that means in terms of the ability to see and analyze more, whether it's in biology or other areas, is exciting and profound. And so we see this as another path that we will continue to execute against.

We're really proud of our own. We think that what you've seen time and time again is the breakthroughs from that team led by an extraordinary leader, Hartmut Neven. Published in all sorts of different places, and we just continue to build success to success with the Willow chip being the most recent.

That's a great question. Not surprisingly, I ask that question all the time as well and have been for some time. You know, I think it's still a number of years ahead. It's starting to point to different applications that we're excited about, but whether we're three years, five years, hard to say. But it's getting closer since I started asking that question.

Well, when you think about it, take something like the human body and the complexity of the human body or anything in nature that's multi-dimensional, the ability to actually crunch data more efficiently, to have better insights, is one area that's very exciting for us to think about and the possibilities that come from that.

It's 200,000 paid rides.

Well, I think if I just step back and talk a bit about Waymo because we are very excited about it. You know, we started on that journey more than a decade ago as well. And the original thesis is that more than a million people die on the road every year in accidents, and if AI can help improve the safety of driving because our Waymo self-driver, the AI does not get tired, it does not get distracted, it stays focused on the road, you've got camera sensors everywhere, we can improve safety, we can help save lives. And that was a really exciting motivator for the team. We've been rolling it out. It's been extraordinary to see the take-up. It's now one of the top attractions in San Francisco if anyone's out this way. But we're also in LA, in Phoenix, and Austin, and expanding. And we're going to continue to expand because you see both the reaction to it. When people get in, some people are anxious about it at first, and then within literally under a minute, they see they just go right into whatever it is they wanted to be doing. There's a safety element around it. So we think it'll continue to be rolled out. We're doing a pilot in Japan right now. And there's an opportunity we think to help save lives and are excited about doing that.

You know, I think it's too early to call. A couple of reasons. One, many people, in particular in the states and other developed markets, own more than one car. There are good reasons to own a car. And so we'll see, human nature and human behavior changes over time. The other is that Waymo is primarily now focused on what are these robo-taxis as you described it, but it's reasonable to assume that this capability, this extraordinary technology, can also be an assist with cars that are owned. And so there are a lot of execution paths which enable you, me, others to decide how they actually want to get around.

So capital allocation, I think one of the core elements in capital allocation is that you need to ensure that you're continuing to invest aggressively for the long run. If you don't invest for the long run, you're sowing the seeds of your own destruction. And that's a lesson I've seen throughout my career over and over. And so very important, it's been core to the ethos obviously of this company since inception. In the early days, there was a kind of a mantra which is 70% in your core, 20% adjacent, 10% moonshots. And that's evolved over time, but this core sense of you've got to continue to invest aggressively for the long run remains core to who we are. And we're at such an exciting time in history given the opportunity with AI, you know, both on the consumer side and on the enterprise side. So we're certainly making investments there. I end up spending a lot of my time globally because what is really key is every head of state is saying the same type of thing, which is I want to be a part of this digital transformation. It is key for all the reasons that we've already talked about. And so what we spend time looking at is, as we're continuing to invest in our technical infrastructure globally, in other words data centers, subsea cables, how do we link up the world? The opportunity is how do we engage more deeply? What work are we doing on their behalf to help accelerate their digital transformation? How do we engage on the public sector side? And so there's a really global view to this.

So the moonshot factory, as it was lovingly named years ago, is one of the core areas within Other Bets. It's called X. And out of X came, for example, Waymo. It was incubated in X and then got moved out to become an independent company. They've also worked on, for example, Wing, or incubated Wing, which is our drone business. Excited about Wing. They're doing work as an example with Walmart, and we're seeing really exciting results there and see the upside. They're also incubating, and you'll hear more about a company called Intrinsic, which is a robotics operating system. And so they have a number of different things that they've been working on. I think one of the very important elements there is that when they are approaching any incubation, their mindset is obviously you can't incubate everything and have it work well, and so kill things fast in order to move into the areas that are the most promising.

Well, we've got everything from kindergarten all the way through to postgraduate robotics.

I think there are a lot of elements that build resilience, and one of the important areas is continuing to invest in subsea cables, but it's also the technical infrastructure more broadly. So it's data center resilience, data center redundancy, so that you can actually be positioned to serve customers when and where needed, and in the event of something that happens, and things unfortunately always seem to happen around the world, you've got the resilience needed to continue to operate at a high level. It's our cybersecurity defenses. One of the things I've learned in my career is it's much easier to prevent than to fix a problem. And so building in this strength upfront is key. And it's technical infrastructure. I can be happy to talk more about cables because I'm really proud of the network we have globally, but it's the other elements like a zero-trust approach on cybersecurity so that you're fortifying defenses in a world that is continuing to become ever more challenging. And I think that's one of the key additional applications of AI that's important.