Sundar Pichai CEO of Google: AI prone to errors

0:00This BBC podcast is supported by ads outside the UK.

0:30Find out more at polandyourplacetogrow.com. If journalism is the first draft of history, what happens if that draft is flawed? In 1999, four Russian apartment buildings were bombed, hundreds killed. But even now, we still don't know for sure who did it. It's a mystery that sparked chilling theories. I'm Helena Merriman, and in a new BBC series, I'm talking to the reporters who first covered this story. What did they miss the first time? The History Bureau, Putin and the apartment bombs. Listen on BBC.com or wherever you get your podcasts. Hello, I'm Faisal Islam, the BBC's economics editor. And this is the interview from the BBC World Service.

1:18The best conversations coming out of the BBC. People shaping our world from all over the world. I'm disappointed in him. Yeah, a deal done four times and then you go home and you see just attack a nursing home in Kiev. I said, what the hell was that all about? I was still in an induced coma in hospital when the world was defining me, but I was still 15 years old and I did not know who I was. I love singing and so my goal was always to do better and better at it. Today we are spending trillions on war and peanuts on peace. For this interview, I met Sundar Pichai, boss of Google, at the tech giant's headquarters in California.

1:56He is the chief executive of Alphabet, holding company for Google and other subsidiaries, which sits at the forefront of the AI revolution, a technology he describes as the most profound that humanity has ever worked on that will change how we live and work. But he also acknowledges the risks that come with artificial intelligence. Some AI models are still prone to errors, he warns, and should not be blindly trusted. There is the impact on jobs, but also on the climate, thanks to AI's immense energy needs. Mr Pichai admits Alphabet's progress towards its target of achieving net zero by 2030 has slowed, but the goal remains in place.

2:37The tech industry, he tells me, is at an extraordinary moment, with trillions being invested in artificial intelligence. Alphabet's shares have doubled in value in recent months, reaching$3.5 trillion earlier this month. But Sundar Pichai warns no company, not even his, would be immune if the bubble were to burst. I look at the actual progress we are making in terms of the model capabilities and the progress is palpably exciting. And given the potential of this technology, the excitement is very rational. It's also true when we go through these investment cycles, You know, there are moments we overshoot, right, collectively as an industry.

3:20We can look back at the internet right now. There was clearly a lot of excess investment, but none of us would question whether the internet was profound or did it drive a lot of impact. It's fundamentally changed how we work digitally as a society. I expect AI to be the same. So I think it's both rational, and there are elements of irrationality through a moment like this. Welcome to the interview from the BBC World Service with Sundar Pichai. It's an extraordinary moment even by Silicon Valley standards. Every decade or so, you have this inflection points. You have a new technology. It was a personal computer at one point, the internet coming in the late 90s.

4:11Then it was mobile. Then it's been cloud, what we call as cloud. Now it's clearly the era of AI. And can you use a sense of the scale? Clearly, the metric that people kind of go to is the market cap, these extraordinary numbers for you, three and a half trillion, NVIDIA at five trillion, the sheer amount of investment going in. Give us a sense of scale. One way to think about the scale is what we are all investing in capital to build out the infrastructure that's needed for artificial intelligence. You know, maybe four years ago, Google was spending less than$30 billion per year. This year, that number is going to be over$90 billion.

4:54dollars. If you collectively add what all the companies are doing, we have well over a trillion dollars of investment going in, building the infrastructure for this moment. And one way I think about it is in the next couple of years, we'll end up building what we probably built in the past 10 to 20 years. So that gives you the scale at which this is ramping up. Now, you mentioned some of those phases of technological advancement that happened with much market excitement as well. And the obvious question is around the whole of this country, around the whole world right now, is it a bubble? Look, two ways of thinking about the question.

5:30I look at the actual progress we are making in terms of the model capabilities, and the progress is palpably exciting. And people are using this. We are deploying it in our products. Consumers are excited about using it. We are giving it to companies. They are using it to make their The company is better. So you see real demand, and we are constrained in our ability to serve that demand. And given the potential of this technology, the excitement is very rational. It's also true when we go through these investment cycles, there are moments we overshoot collectively as an industry. We can look back at the internet right now.

6:08There was clearly a lot of excess investment, but none of us would question whether the internet was profound or did it drive a lot of impact. It's fundamentally changed how we work digitally as a society. I expect AI to be the same. So I think it's both rational and there are elements of irrationality through a moment like this. You just described your excitement about Google's kind of hardware, which is the key scarce resource. Because I guess, would your argument be, I guess it would be, that whatever happens, whatever appears to be a frothy, exotic deal from one of your competitors, Google will be immune from any bursting of the bubble because of your spectrum of investments across all technologies.

6:55No company is going to be immune, including us. If you over invest, you know, we'll have to work through that phase. But, you know, we are, I think we are better positioned for many years, we have taken a deeply differentiated approach to how we've approached AI. One of the first things I did as CEO was to shift the company to what I called as an AI first approach. And part of that was doing all the parts of what it takes to build great AI technology. So we call it a full stack approach. You can think of it as a soup to nuts approach, all the way from the underlying physical infrastructure to the research that you need to do to drive this technology forward to deploying it in products and platforms, be it search, be it in YouTube, be it in Android, and so on.

7:45And so we have taken the deep approach. So we are able to invest at scale and make it work across all these products and businesses. And I think that we are in a better position to take a long-term view and approach this moment. Let's take that and then let's try and sum up, if you like, the ultimate power of the tools you have built for people at home. How effective could it be? Could an AI agent do your job at some point? Look, I think where we are is right now you can interact with this AI, ask questions, go back and forth, and have these intelligent exchanges on many, many topics. I think the next step in the next 12 months, you will see they are able to do more complex tasks for you.

8:27So that's where it gets really interesting. You know, I have to go shop something. I have to buy a birthday gift for my spouse. And, you know, can I ask this chatbot to go do that? Right. So that kind of what we call agentic experience, this is what we are all excited about. Down the line, that means there are moments it can help you make a decision. Right. Like, you know, it could be, should I invest in the stock or my doctor is recommending a treatment? And how should I think about the pros and cons of the treatment? So those are all real tangible use cases. There is still work to be done to unlock those capabilities.

9:02But that's the journey which has been so exciting to see. OK, so still CEOs are safe in their jobs. I think what a CEO does is maybe one of the easier things to maybe for AI to do one day. But the whole point of the value and productivity kind of offer to companies that are buying all your goods and your services is to automate many human tasks, is it not? Let me put it this way, right? I think people today are juggling many things, right? And people are overloaded. We've always had, back in the history, you know, it could be a dishwasher coming to your home. I remember growing up, you know, when we got our first refrigerator in the home, how much it radically changed my mom's life.

9:49You can view it as it automating some, but, you know, it freed her up to do other things, right? So let's take the example of a radiologist. The number of scans people are getting is growing year on year. And the number of images per scan is also rising pretty significantly. How do you help an audiologist cope up with this increased demand? Maybe an AI tool can help that way. So I think that's what you will see more or less. I get it. You want to focus on the win-wins, but it's many people, many aspirational middle classes across the West are thinking, hang on a minute, it's affecting lawyers' jobs, it's affecting creative industries, it's affecting accounting jobs, it's affecting journalism too, it's affecting the media in general.

10:32So, and they're surprised by it. And they're wondering, do you know which jobs are going to be safer? Have you got an idea? Two things. One is, first of all, look, I said many years ago, AI is the most profound technology humanity is ever working on. It has potential for extraordinary benefits. And we will have to work through societal disruptions. And you are highlighting, you know, it'll end up creating new opportunities. It will evolve and transition certain jobs. And people will need to adapt. And then there will be areas where it will impact some jobs. So as a society, I think we need to be having those conversations.

11:10And part of it is how do you develop this technology responsibly and give society time to adapt as we absorb this technology? I think these are all very, very important and fair questions. Just very practically, parents don't know what to advise their kids anymore, just to ride out this airwave. Give us some tips. Based on what I see, I won't change anything of how we've always thought. I think there's going to be a wide variety of disciplines which will end up mattering. I would encourage the next generation to embrace the technology, learn to use it in the context of what you do. And I think people who will learn to adopt and adapt to AI will do better.

11:52It doesn't matter whether you want to be a teacher or a doctor. All those professions will be around. but the people who will do well in each of those professions are people who learn how to use these tools. All of the hopes, the hype, the valuations, the social benefit of this transformation you've just described, they're built on a central assumption that the technology functions, that it works. Let me propose one simple test. Of Gemini, which is your booming chat GPT kind of competitor, is it accurate always? Does it tell the truth? Look, we are working hard from a scientific standpoint to ground it in real world information.

12:27part of what we have done with Gemini is we have brought the power of Google Search. So it uses Google Search as a tool to give answers more accurately. But there are moments, these AI models fundamentally have a technology by which they're predicting what's next, and they are prone to errors. Yeah. You know some of the examples, an example of glue as a pizza ingredient, a sitting senator wrongly accused of assault. I mean, this is bad, isn't it? I think we take pride in the amount of work we put in to give as accurate information as possible, but the current state-of-the-art AI technology is prone to some errors.

13:04This is why people also use Google Search and we have other products which are more grounded in providing accurate information, right? But the same tools are helpful if you want to creatively write something, right? So you have to learn to use these tools for what they're good at and not blindly trust everything they say. Okay, don't blindly trust. But let me suggest to you that you have a special responsibility because this whole model, transformer model, the TNCHAT GPT, was invented here under you, and you know that it's a probability. And I just wonder if you accept the end result of all this fantastic investment is that information is less reliable.

13:41If you only construct systems standalone, you know, and you only rely on that, that would be true. Which is why I think the information ecosystem has to be much richer than just having AI technology being the sole product in it. Truth matters. Yeah, truth matters. Journalism matters. All of the surrounding things we have today matters, right?

14:08You're listening to The Interview from the BBC World Service. People shaping our world from all over the world. This advertisement feature is paid and presented by Poland's Ministry of Foreign Affairs. Looking for the perfect place to grow your business? For over 30 years, Poland has been a launchpad for innovation at the heart of Europe, powered by world-class research and ambitious young talent. From bold startups to established powerhouses, Poland offers the momentum, resources and reliability you need to take your business further. Poland, where businesses don't just grow, they're built to last.

14:42Find out more at polandyourplacetogrow.com. If journalism is the first draft of history, what happens if that draft is flawed? In 1999, four Russian apartment buildings were bombed, hundreds killed. But even now, we still don't know for sure who did it. It's a mystery that sparked chilling theories. I'm Helena Merriman, and in a new BBC series, I'm talking to the reporters who first covered this story. What did they miss the first time? The History Bureau, Putin and the apartment bombs. Listen on BBC.com or wherever you get your podcasts. For this episode of The Interview, I'm speaking to Sundar Pichai, chief executive of Google.

15:28He's not a showman like Elon Musk of Tesla or the black leather jacketed Jensen Huang of NVIDIA. He's fairly softly spoken, cerebral, geeky even. And I felt that this was a very candid conversation. CEOs are getting used to not answering questions from actual journalists and answering them from their fellow CEOs or investors. Only a year ago, Google was seen as under threat. Its essential business model of search being challenged by ChatGPT, its share price actually not enjoying the same sort of booms as some of the other big tech giants. And I think he feels some vindication right now about Google's share price and the appreciation for some of its technology because it certainly wasn't there a year ago.

16:12OK, let's return to my conversation with Sundar Pichai. The scale of the AI build-out is creating another trade-off, not just for you, but for humanity, on energy. I think this is an opportunity to be frank, Mr. Pichai, with the world. Is there a new calculus now? Is the build-out of AI more important than climate? Over time, I don't think, you know, this doesn't need to be a trade-off or a zero-sum game. I think one of the things I'm excited about, because the energy needs through this transformation is so immense, We, as well as others, we are investing to develop new sources of energy. We just finished signing the largest corporate purchase for nuclear fusion energy.

17:00We have many purchase agreements for energy from small modular nuclear reactors. We're using geothermal energy in our data centers. So the amount of dollars, R &D dollars, capital investments going in these new sources of energy, I think will actually accelerate. So you're right, AI is dramatically increasing demand for energy in a way that the current systems can't fully cope up. But that is driving extraordinary investments in solar, in battery technology, in nuclear technology and other sources. So I think, you know, I am, as a technologist, I'm optimistic through this moment that we will have abundant sources of renewable energy in the future.

17:41I mean, by the end of the decade, I think more energy will be used by data centers than the whole of India, 50 % more than all of the EV fleet. You've sort of parked or dropped your 2030 sustainability net zero target? No, we still have it. We publish progress reports against it. You are right. some of the progress, the rate at which we were hoping to make progress, will be impacted because we are seeing a much faster than expected growth in the underlying build-out. But we are meeting the moment by investing in all these new technologies. And so I think that's how we are trying to make it work.

18:19Another key fuel for the AI boom is obviously the content that it's trained on. And the tech companies, including Google, have relied on fair use and have sort of scraped books, music, journalism, and are sort of selling back some of that expertise to the world. Do you accept that companies like Google will have to end up paying in some way for that? Throughout, first of all, to step back, I think it is so important as we go through this that we both help drive creativity and innovation, but we have to do that in a framework which respects creators' rights, as well as a love for transformative use to deliver benefits to society.

19:00I think we are committed to copyright frameworks in all the countries we operate in. Today, when we train, we give people an opportunity to opt out of the training. And we honor copyright in terms of how our outputs are generated. And we are in the process of working with the industry to create newer frameworks as we move through it. and, for example, in YouTube, we've always incorporated an approach to deliver value back to content rights holders. We will apply those same principles through this AI moment, and I think it's super important to do that, and so we're committed to getting it right.

19:40Yeah, in the US, it's basically been settled in the courts. In the UK, in Europe, there's a debate about whether legislation is necessary. You have famous pop stars like Sir Alton John saying it's thievery on the high scale. If someone wants to use my song, ask and have transparency in how it's used. Are you willing to do that? Today, we allow anything we train, you know, people can choose whether their content is opted into their training or not. So we do give people those rights. Oh, I want to take you back to the beginning of this year, to that totemic kind of tech bro photo of Zuckerberg, Bezos, Musk, and yourself in the seats of honor at the president's inauguration, you're already powerful.

20:20You have access to what is now the world's most powerful technological tool, and now not being maybe counterbalanced by political power, but some would say joined at the hip. Can you see why that makes many people uncomfortable? Look, I think if I were to speak about the US, it is an extraordinary moment in terms of this AI technology. I think it has tremendous opportunities to deliver benefits to the economy. It matters from a national security standpoint. As one of the leading technology companies, which is pioneering and driving progress in AI, we are deeply committed to engaging constructively in the U.S.

20:55President Trump has been very clear in the importance of this technology and has articulated a clear AI action plan with multiple aspects to it to deliver benefits to the country. So we are deeply engaged as a company of our scale should, and not only in the U.S., but with other governments around the world. I think this is an important moment, and not just us, the leading companies which are building this technology. I think it's important we are engaging with the government in a way that this technology is going to translate to deliver benefits for society, including thinking through concerns around potential misuse of the technology.

21:31So I think we are going to need deep industry-wide frameworks. As a company, we want to engage with the governments, with nonprofits. profits. I think you have to bring in all stakeholders into this. On a specific issue, what will the impact be of the White House's crackdown on foreign worker visas on your company and the sector? How do you feel personally as someone who arrived on an H-1B visa? Look, if you go look at many of the fundamental breakthroughs, many of them are immigrants. If you look at the history of technology development, the contribution of immigrants to the sector has been nothing but phenomenal.

22:08But I do think the government understands it. I think there's a framework by which we all can still bring talented individuals. I think they're making changes to address some of the shortcomings in the current program. And I think we'll be able to continue investing. Okay. Well, the last time the BBC was here, a few years ago, you had already invented the core technology that would go on to become ChatGPT and Gemini. And at that time, no one had really heard of it outside of Google. And yet it's transformed the world quite clearly. Is there anything similar you've got buried away in the labs with the white coats now that we should know about?

22:45Look, I mean, we are working on a range all the way from, you know, our self-driving technologies making extraordinary progress, and it's already showing safety benefits. And, you know, I think scaling up the technology will avoid a lot of human fatalities, which I think has a lot of societal benefits. I genuinely think what we call in this umbrella as AI. You know, we didn't even talk about things like AlphaFold, which won the Nobel Prize, and which is helping. There are many, many biologists and chemists around the world using it to better discover new drugs. Is this right? You did hundreds of millions of years of PhD research in a few months?

23:25That's right. That's exactly right. It would take one PhD, the entire PhD, to do one protein. And we have done around 300 million proteins in a matter of a few months, and we've made it openly available to everyone. I think in some ways we can take progress here for granted. We all used to talk about the Turing test. We've kind of gone past it and no one talks about it. If I had told five years ago to people that if you go to San Francisco, there'll be many driverless cars driving around without anybody in the driver's seat, people wouldn't have believed it. It just happened. Hang on. Didn't they start honking at each other?

24:00Were they talking to each other? Were they communicating? I hope not. I don't quite think so. Are they alive? Maybe. Look, we are using AI to better understand how dolphins communicate, et cetera. And so I do think there'll be serendipitous moments like that. But it's exciting. And what is important is, as humans, we adapt to these technologies. The fact that I recently had my dad in a Waymo car, he is over 80 years old, the wonder, you know, I was in the back and he was in the front, seeing him experience it, you know, helped me understand the progress we all take for granted. And I think there are going to be many such wonderful things in the future.

24:46Thank you for listening to the interview from the BBC World Service. You'll find more in-depth conversations on the interview wherever you get your BBC podcasts, including episodes with former US Vice President Kamala Harris, cycling champion Sir Bradley Wiggins and Russian punk activist Maria Aljokina. Until the next time, bye for now.

25:26Poland has been a launchpad for innovation at the heart of Europe, powered by world-class research and ambitious young talent. From bold startups to established powerhouses, Poland offers the momentum, resources and reliability you need to take your business further. Poland, where businesses don't just grow, they're built to last. Find out more at polandyourplacetogrow.com. If journalism is the first draft of history, what happens if that draft is flawed? In 1999, four Russian apartment buildings were bombed, hundreds killed. But even now, we still don't know for sure who did it. It's a mystery that sparked chilling theories.

26:06I'm Helena Merriman, and in a new BBC series, I'm talking to the reporters who first covered this story. What did they miss the first time? The History Bureau, Putin and the apartment bombs. Listen on BBC.com or wherever you get your podcasts. If you knew that, given me your return now on my birthday, I wouldn't mind. A rebuildinghead is called on Nutrients, so it being opened at many sites.