Delighted for our next event, which is a fireside chat moderated by Anaf Kareim, featuring Ruth Porat. I'm going to introduce Anaf, and then Anaf is going to introduce Ruth. Anaf is the founder and managing partner of Latitude Capital, a VC firm focused on software and artificial intelligence. He has a long list of accomplishments, including being a board member of the San Francisco Marin Food Bank, a member of the Council on Foreign Relations, a graduate of Stanford University, where he served as senior class president, and he holds an MBA and MPP from Harvard. Please join me in welcoming Anaf and Ruth to the stage.
Thank you, Neil, for that kind introduction. Ruth, good to have you here and good to see you.
Great to be back on campus. Ruth doesn't need an introduction for any room she walks into, and certainly not here at Stanford. But for some context, Ruth grew up in Palo Alto, her father was an engineer at SLAC, the Stanford Linear Accelerator. She did her undergraduate work at Stanford, she was on the Board of Trustees at Stanford and on the board of the management company at Stanford. Outside of her Stanford service, she's had an okay career as well. She has a graduate degree from the London School of Economics, an MBA from Wharton, and started her career on Wall Street at Morgan Stanley, where she served in various leadership positions including CFO. Then about a decade ago, she decided to trade in the gilded halls of finance for sunny California again and came back and became CFO of Alphabet, and now holds the title of President and Chief Investment Officer of Alphabet. So thank you again, Ruth, for being here.
So Ruth, every day we wake up and read about AI in the news, not just in Silicon Valley but everywhere. I want to start there. Alphabet has said they are investing $75 billion just this year in AI and building out infrastructure and capacity. If we rewind the clock a quarter century ago, Google transformed the way all of us engage with technology and the internet. If we want to find a piece of information, we used to have to go to Green Library to find out what the GDP of Norway might be, but today we can just do it with a click of a button. If we fast forward five years from now, what is the promise of AI for all of us sitting in this room? How will it impact our daily lives in a tangible and practical way?
I think we are living in one of the most extraordinary times as a result of precisely your question, the potential with AI. When we talk about it, we think of it in four ways. First is the economic upside. Estimates in the US alone are that AI could add $4 trillion to GDP by 2030, and could add $20 trillion annually to global GDP in a decade. That assumes what economists call diffusion, really implementation across industry and across the public sector. It's not about using a chatbot; it's about radically rethinking what is your business, how do you interact with customers, how do you think about all elements of operations, how do you think about risk management. I keep calling it a radical rethink of what is your business, and there's tremendous upside that comes with it. For each one of us, starting on that journey is important. It is still very early days to radically rethink one's business. The other thing that's really exciting is that AI is advancing science. I'm really proud to work with two Nobel Prize winners who just won the Nobel this past fall, Demis Hassabis and John Jumper, for something called AlphaFold, which hopefully many of you are very familiar with. Simply stated, it has been described as the single greatest contribution to drug discovery in our lifetime. They chose to open source it, and two and a half million scientists around the world are using it in 190 countries. As Demis describes it, this is science at digital speed, addressing solutions for diseases. The third is everything around critical social issues. What can we do to enhance education, provide operating leverage for teachers, individualize paths for students? A lot of people in this room can speak about this more eloquently than I can. There's also what we can do with healthcare, early diagnosis, better care. So I'm super excited about social issues. And then something I'm looking forward to hearing the next panel talk about is the application to security and fortifying security. The potential across every one of those areas is profound, and we're at a really early stage. To answer your question with one more story, at Google 20 years ago we started on this journey with Google Translate. Over the last 20 years, we've translated 250 languages. In the last six months alone, we added 110 of those 250 languages. The reason I wanted to end with that is it's probably the best way to answer your question. You move along at a certain pace, and then it inflects, and that's where we are with the application of AI. For each one of us, thinking about how more profoundly we can think about whether it's a business or the way we approach our discipline or how we run a city, it can be pretty profound.
If we take what happened in November 2022, when most of us in this room probably had the first chance of actually engaging with AI, that was the ChatGPT moment. Most consumers probably understood for the first time how they could engage directly with AI and leverage it for their own capacity. You and I know that Google has been at the forefront of AI for many years before that, including the foundational research that led to a lot of those advancements. I wanted to ask you, if we put ourselves in the shoes of Google and the management team at that time, even though AI had been in the ads business, behind the scenes in Google Photos, Google Lens, and so many other products, what was the decision around not releasing a direct-to-consumer product that would really capture the zeitgeist in the way that Gemini has now, but ChatGPT did beforehand?
It's an important question. Our view very strongly is that Google Search is about trust and quality. You expect accuracy when you put your query for whatever it is you want. Sundar talked about this quite a bit. It's like if your child is sick and you wake up in the middle of the night and you Google for 'my four-year-old, what's the appropriate dose of Tylenol?' It can't be wrong. Many people learned the term 'hallucination' when ChatGPT broke out to the forefront. We knew hallucination was a risk and weren't prepared to take the risk with something as important as Google Search. We found a way to weave it together with AI Overviews so that you are playing with it and have that firsthand experience. For us, what's really exciting is building on this momentum we see. Our view very strongly is that it is still very early days, but Google comes at it with a very differentiated approach. You start with the research that we have led, by Nobel Prize winners, and then to your point, it was experienced and touched and felt by people through ChatGPT, but they had already been living it for so many years. Billions of people are already benefiting from Google AI, whether it's something like autocomplete in email or your photo search, things that we take for granted. We don't think of that as AI; that just is a product working better. It goes right into the chips that we use and the ability to deliver models. We're accelerating the pace of what we're seeing. But taking you back to that moment, reputation is so important in delivering for people around the globe. That was very important and is important.
Talking about safety, risk, security, I want to talk about DeepSeek. For the audience, last month there was big news around a Chinese company releasing a model called DeepSeek, which many of you probably read about. I think it shattered a lot of perceptions in the AI community around a couple of core assumptions, both across the fact that maybe the ROI or the necessity of investment across AI could yield a higher ROI with lower investment, and the fact that China is probably not that far behind us in AI as many experts had deemed to be the case over the last couple of years. Curious if you can use that as a jumping-off point to talk about China and the US as it relates to AI competition. What did you see in the DeepSeek moment? How has that impacted the way you think about US policy as it relates to competition in AI?
There's a lot in your question. I think the one thing that we and others have been very focused on for quite some time is increasing efficiency of every element of compute. It's efficiency in models, efficiency in chips, and that's been a key journey that we've been on. We saw it in what we call classic compute, and we're seeing it as well in AI compute. Our most recent TPUs, which are basically our version of GPUs, are 67% more efficient than the prior generation. Models are more than 100 times efficient. Each one of us who is working on this is powering that efficiency curve at a very rapid rate, and that's in part what you're seeing with DeepSeek. Our view is still that the US companies are ahead when it comes to frontier models and chips, but I think it was a very helpful moment in certain respects because there's clearly a tremendous amount of innovation and focus in China. Leadership never can be taken for granted, as anyone sitting in Silicon Valley knows. This put a real spotlight on the fact that the US may be in the lead now, but don't take it for granted. We want to ensure collectively that we have a pro-innovation investment environment that still protects on the downside, but we can't take this for granted. That was one really important element of the experience.
You talked about this in relation to the US and China when we spoke a couple days ago. You were at Davos last month, you were in DC earlier this week, you've been speaking with a lot of world leaders about AI, and it sounds like the refrain you keep hearing is that no one wants to be left behind in the future of AI. Tell us about the policy implications for the US. How should we be positioning ourselves as it relates to the broader global community?
It has been a recurring refrain traveling the globe. Heads of state say they see this moment, they see the economic upside, and they also want to talk about what it means for the delivery of education or healthcare, or in certain places, how can we enhance agriculture, the lowest strata of the economy. What can we do to enhance productivity yield? So wherever I go, there's a sense that this is a moment in time and we can't be left behind. What I also very clearly hear is just this deep respect for the US and US values, wanting to work with us. But if they can't work with the US for whatever reason, they're not going to be left behind; they will find an alternative approach. So I think levels of engagement continue to be important, ensuring again that that's the approach as we're collectively looking at what enables all of us to capture the upside while protecting on the downside. There's a really current issue in front of the administration now, which in the final days of the prior administration they came out with some export controls called the AD Fusion Rule. It limits exports of chips and also classifies our allies in tiers, and that's been befuddling for many because the question is, do we want to engage or do we not? I think there's a journey we need to go through that maximizes the opportunity to engage while putting a lot of the safety guardrails in place, because I don't think it's appropriate or responsible to suggest just throw it overboard. Those are the types of things where it's imperative that we have the opportunity to help build out whatever is needed as countries are going on their journey.
Are you concerned about the future state of a world where there's more of a haves and have-nots as it relates to AI power and the economic capture related to that? How do you think about that in that context?
That's why I think it's such an important question. When I hear over and over again, 'We cannot be left out of this transformation, we see the economic upside,' it's fascinating to me the number of times the head of state doesn't actually start with 'Will you build a data center? What kind of dollar commitment will you make to our country?' I've often heard, 'What will you do to help train our people? What can you do on education?' Time and time again, hearing similar refrains, I came up with a visual model of a triangle that's incredibly important for every country. What's in the triangle? First, it starts with Google's mission: organize the world's information, make it universally accessible and useful, which simply put means make sure that everyone everywhere can have the same access to information, products, and markets that everyone here in Silicon Valley is privileged to have. How do you get that? First, you need to have the infrastructure in place. You need data centers, and very importantly, you need to be linked into this digital world. We build subsea cables around the globe; that's an important part of the unlock. Very much to your question about this divide, as an example, we built a subsea cable from Africa to Australia. When you do that, you're building this big trunk, and you can then light up countries and island nations along the way. When we connected Fiji, the Prime Minister said, 'This is a gift to all of Fiji because it creates so many opportunities, it ensures that everyone has this access.' That's what's really exciting. That's the second part of the triangle. It's really important for all of us to keep coming back to the fact that one-third of the globe is still not connected. So you need to be connected. Then it goes to the third part, which is, what about education and building a sense of what is that opportunity, and very importantly, what are the skills that you need to thrive in this digital economy? That's the triangle. I think only when we come at it thinking about all points of the triangle can you get to your question: can you actually close that opportunity gap and help address some of the disparities that we already see?
One point of that triangle is obviously data infrastructure, and a lot of what powers that is energy. I was in conversation with a chief product officer of one of the top LLM companies who said the biggest bottleneck they see for further advancements in AI is actually energy. I know we talked a little bit about the opportunity there, both from the private sector as well as potentially from public sector investment. Unpack that for us a little bit. What are the needs from an energy perspective? How should the US be thinking about that from a national perspective?
First, the context, and you set it up well. The only reason we need more—not the only reason, one of the important reasons we need more energy—is because of this extraordinary opportunity. If implemented well, if there is diffusion across the economy, you have an economic uplift. The other drivers of the need for energy are additional onshoring of manufacturing, electrification across industries. All of these are economic catalysts for the US. So in many respects, we need to invest because we have this unlock opportunity, coupled with the fact that there's been underinvestment in the grid for the last couple of decades. What's encouraging is there are some immediate opportunities that collectively we need to focus on. One is that the way the incentives around the grid work, you are incentivized to add capacity, not optimize existing capacity. According to the prior Department of Energy, if what are called grid-enhancing technologies—things for example to load balance across the grid at different times as needed—are applied to the grid, you can unlock up to 100 gigawatts of capacity just with existing infrastructure. Having lived my whole life in the private sector, you wouldn't add capacity before you optimize what you have. You're not going to maximize return on invested capital if you're not optimizing what you have. So that 100 gigawatts of capacity by applying grid-enhancing technologies is exciting. At Alphabet, we've been incubating a company called Tapestry for about seven years now, which is all about grid planning and the ability to actually help with some of these issues. We've been working with the grid in Chile and have been able to take some decisions down from weeks to days. That kind of work is critically needed. Then there are some things around permitting to get this backlog onto the grid, but there also is some very important work around adding additional sources of energy. One of the things we've done at Google for a long time is forward market commitments, because we find that actually ends up catalyzing more investment. For example, we were early in geothermal with a company called Fervo, which we're really proud went through development stage and is now moving on to the grid. We're the first corporate purchaser of SMRs, small modular reactors, on nuclear with a company called Kairos. What was important about that is it's multi-tranche, like five to seven tranches, so we can keep building these small modular reactors. These are going to be important additions to what's being done. There's obviously a lot of capital flowing in, and you're seeing more and more interesting structured solutions. We did one with TPG here in San Francisco and Intersect Power to look at building campuses where you bring together the energy sources with what you need for data centers. So a lot of focus on it. Definitely need to expand capacity, but I started where I did because that's a good thing. If we didn't have growth, we wouldn't need it. I'd rather have the growth and put the investment behind it.
The third leg of that triangle is on education. I want to touch on that. Obviously it's an international issue, but even here in the US, if we think about the downstream impact of AI, for the first time we've gone through technology revolutions before—the Industrial Revolution, we've seen impacts of globalization have a huge impact on workforce and labor. But this one feels a little different. There's an opportunity here where the diffusion of technology changes the pace of change, as well as the fact that for the first time, the people being affected might look more like knowledge workers, white-collar workers. We're sitting here at Stanford where the number one major is computer science. We're seeing those safe areas, professional areas, getting potentially disrupted. Talk us through how you think about the disruption happening potentially in the workforce, and then how should educators be rethinking some of the opportunities around AI?
Throughout history, more jobs have been created than destroyed. In fact, 60% of the jobs that exist today did not exist 80 years ago. That's just this constant evolution. The way I think about that is, in the aggregate, that's beautiful, but if you're the person who's displaced, it becomes very binary very quickly, and that does not work. So we and many others have put real money and effort behind workforce development programs, ensuring that people have the skills needed in a digital world. What are AI skilling programs? We created an AI Opportunity Fund. I think this becomes something that not one company can do alone; many companies need to work on it. We created a program with a consortium of 250 companies that will take our certificates and look at them as creating the ability to then move into a role in cybersecurity or other areas to enhance this kind of evolution of what is work. I think the public sector has a role to play in that as well. One of my favorite books is actually Carol Dweck's book, which hopefully everyone in this room has read because she's such an extraordinary Stanford scholar, called 'Mindset.' It's all about the importance of having a growth mindset. I've read it a couple of times. I think that at a time like now, it's a reminder that it's important that we do have a growth mindset, because the only thing that's constant is going to be change. So we need to protect again with programs that enable people to evolve. There will be disruption, as there typically is, but that's the effort to smooth it and create new opportunities.
Related, at Stanford, generally in academia, technology has had a tight symbiotic relationship. Obviously Stanford and Silicon Valley, we're seeing something interesting probably for the first time where academics and researchers are entering the private sector with regards to AI at a pretty fast pace. Given the fact that capex investment and access to compute is so important to pushing the ball forward with regards to AI research, what is the role of universities and research centers in this new wave?
I think a critically important role, as academia has always had. We're really proud of the work that we do with universities across the country and around the globe. A lot of it ends up being partnerships with scientists within academic institutions. That's a core part, and oftentimes it's in areas that leverage AI but in related areas, for example in biology, other avenues within healthcare or the sciences. That's a way to bring the best together to try and make advancements. Then we look at different programs, to your point about compute that we can make available to universities. So there's a lot of collaboration that goes on.
One topic which is not part of the triangle but I know is close to your heart is healthcare and technology. You touched on it a little bit in the beginning, but you're a cancer survivor, I know you're deeply involved with the medical community. Tell us about what you're so excited about with regards to healthcare and technology, and why is it that you see a lot of opportunity there right now?
A lot of our engineers are really excited because if you can use technology to do something like save lives, that gets people pretty excited for the obvious reason. It is personal for me. I think what's extraordinary about cancer is that if you can have early detection, you can really make a difference between survival or not, or the course of treatment. Is it stage one or stage two, or is it stage four because you've missed it early on? What's pretty stunning is that 40% of people in the United States will be diagnosed with cancer at some point in our lifetime. The way I think about it is, if you look at the person to your left and the person to your right, statistically one of you will be diagnosed with cancer at some point in your lifetime. At Google years ago, we pioneered early diagnosis of metastatic breast cancer. What's stunning in the testing of it is they looked at a large sample size, and 20% of the cases were missed when manually done. Then when tested when you combine a radiologist with AI, it captured and there were no false positives. The notion that we can actually improve early diagnosis is extraordinarily inspiring. It's operating leverage for doctors, for radiologists. That's true with other forms of cancer. We're doing it with what's called diabetic retinopathy blindness, which comes from diabetes, a big problem around the globe. We're seeing early detection on other diseases as well. When you can use AI to make a difference like that, it goes back to your other question about what about the gaps between those who have and those who don't. This means that average doctors are now going to become superhuman again because they're going to have the assist that comes from early detection with AI. I think that's super exciting. AI is not a panacea; you still need to have the rest of the ecosystem work so you can take care and deliver, but it is an important part of the puzzle. We see this in so many areas with healthcare, so it is to us one of the breakthrough areas.
