James Manyika on global AI and inclusion

How can people in all corners of the world participate in the AI revolution and unlock benefits in their own lives? James Manyika, Senior Vice President of Research, Technology & Society at Google, joins the show to discuss global access to AI and the impact on capitalism, the economy, science, and the workforce. He speaks to Google’s latest AI developments, including tools to warn people about natural disasters like floods and wildfires, along with the virtual research assistant NotebookLM. Read the transcript of this episode here: www.possible.fm/podcasts/manyika For more info on the podcast and transcripts of all the episodes, visit https://www.possible.fm/podcast/ Topics: 01:18 - Hellos and intros 02:36 - James’s career and path into AI 03:59 - UN involvement and diversity of perspectives in global AI 08:04 - Learnings from UN High-Level Advisory Body on AI 10:01 - NotebookLM demo 12:06 - Specific uses of the virtual research assistant NotebookLM 15:03 - Global positive impacts of AI 18:38 - AI and capitalism, the economy 24:10 - AI and climate change 29:17 - Humanity and AI development 32:14 - How to mitigate risks 35:02 - The rising importance of humanist disciplines 36:50 - NotebookLM rollout 39:31 - Truth and bias in the age of AI 40:58 - Rapid-fire questions Select mentions: NotebookLM - https://notebooklm.google.com/ Governing AI for Humanity, from the UN's High-level Advisory Body on AI - https://www.un.org/sites/un2.un.org/files/governing_ai_for_humanity_final_report_en.pdf Project Green Light - https://sites.research.google/greenlight/ “Applying AI to Rebuild Middle Class Jobs” by David Autor - https://www.nber.org/papers/w32140 “Accelerating Climate Action with AI” from BCG – https://web-assets.bcg.com/72/cf/b609ac3d4ac6829bae6fa88b8329/bcg-accelerating-climate-action-with-ai-nov-2023-rev.pdf Possible is an award-winning podcast that sketches out the brightest version of the future—and what it will take to get there. Most of all, it asks: what if, in the future, everything breaks humanity's way? Tune in for grounded and speculative takes on how technology—and, in particular, AI—is inspiring change and transforming the future. Hosted by Reid Hoffman and Aria Finger, each episode features an interview with an ambitious builder or deep thinker on a topic, from art to geopolitics and from healthcare to education. These conversations also showcase another kind of guest: AI. Whether it's Inflection’s Pi, OpenAI’s ChatGPT or other AI tools, each episode will use AI to enhance and advance our discussion about what humanity could possibly get right if we leverage technology—and our collective effort—effectively.

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:01Speaker 00

  Members of my team in Google Research began as a small pilot about a year and a half ago, using AI systems to predict floods. We started this in Bangladesh. It worked, and worked in the sense that you can give people up to at least five days advance notice, which makes a huge difference. And so we went from that to where today, we're now doing this in over 80 countries.

• 0:33Reid Hoffman?

  Hi, I'm Reid Hoffman.

• 0:35Aria Finger?

  And I'm Aria Finkler.

• 0:36Reid Hoffman?

  We want to know what happens if, in the future, everything breaks humanity's way. What we can possibly get right if we leverage technology like AI and our collective effort effectively?

• 0:48Aria Finger?

  We're speaking with technologists, ambitious builders, and deep thinkers across many fields: AI, geopolitics, media, healthcare, education, and more.

• 0:59Reid Hoffman?

  These conversations showcase another kind of guest, whether it's Inflection's Pi, OpenAI's GPT-4, or other AI tools. Each episode, we use AI to enhance and advance our discussion.

• 1:11Aria Finger?

  In each episode, we seek out the brightest version of the future and learn what it'll take to get there.

• 1:18Reid Hoffman?

  This is possible.

• 1:25Reid Hoffman?

  AI is throwing a mirror back at us. That's how our guest today has described the way AI technologies and the resulting power and capabilities get integrated into society.

• 1:38Aria Finger?

  When it comes to game-changing new technology, the first waves of people to gain access and benefit are, more often than not, extremely limited. We've seen this rollout pattern take place throughout history with the advent of electricity, cars, and even books. And as we've discussed on the show before, with everyone from Van Jones to Microsoft's Kevin Scott to Khan Academy founder Sal Khan, this does not have to be the trajectory with AI.

• 2:02Reid Hoffman?

  Today, we're asking, how can big tech and business actively support making AI accessible on a mass scale as opposed to a select few countries and companies?

• 2:13Aria Finger?

  And here with us to explore this is James Manyika, the Senior Vice President of Technology and Society at Google. His experience spans big tech and business, along with advisory roles at the UN and the White House. We talked to James about designing and leveraging AI for citizens worldwide, the economic and scientific opportunities associated with this, and yes, some of Google's latest AI news and advancements.

• 2:36Reid Hoffman?

  Here's our conversation with James Manyika.

• 2:47Reid Hoffman?

  I've been waiting for this for a number of months. Welcome to the Possible Podcast.

• 2:54Speaker 00

  Well, thank you, Reid. I'm so excited to be here with you and Arya.

• 2:59Reid Hoffman?

  I've had the pleasure of getting to know you through our work on the Stanford Human-Centered AI Institute. And I think it's fair to say that your journey into AI and your work in the field is, you know, somewhat unusual. So can you share how you came to AI, along with what you're most focused on today?

• 3:16Speaker 00

  The first thing I ever published in my whole life was actually a paper on AI in 1992 as an undergraduate. It was actually on modeling and training neural networks. This is in Zimbabwe, which also surprises people sometimes, given the timeframe. I went on to do a PhD in robotics at Oxford. And from there, I also worked at Jet Propulsion Labs, actually, in the Man Machine Systems Group, when we were working on some of the machine learning systems for the pathfinder programs. I'm Detorit McKenzie, but I stayed connected to AI. In fact, I was involved with my friend Demis Hassabis at DeepMind quite early on. Yeah, now I'm at Google.

• 3:59Aria Finger?

  We talk a lot about AI in the context of global development, and you're involved with UN's High Level Advisory Board on AI. Some people say like, this is what we need. We need a high-level global body to be sort of overlooking AI and regulating it. Other people have questions about the UN as sort of the most effective player. Can you tell us more about like what that looks like, your involvement in it? And what insight has it given you into global ideas about AI?

• 4:27Speaker 00

  Well, that's been a very fascinating time serving on that high-level body, which I have the honor of co-chairing. We have 39 members from 33 countries. So you can imagine very diverse perspectives. And members come from academia, some society, private sector and so forth. But here's the thing. One of the things that's made me realize is just how different perspectives and views are about AI from around the world. And if I were to make a big general summarization, I think I would say people in the global south, the so-called global south, generally are more positive on AI as a way to transform societies, improve the world, address big societal challenges.

• 5:12Speaker 00

  However, however, they do have some questions and concerns. And generally, those tend to be of two kinds. One is they don't feel like they've been as involved in the development of AI or any of these discussions about governance and all the rest of it. And then second, the challenge of capacity, meaning even as basic things as broadband access, connectivity and so forth, which should allow them to participate. So it's a very complex picture. The question about the UN itself, one thing I've come to appreciate is that the UN has zero power. Other than what the members give it and the members of the hundred ninety three countries.

• 5:55Speaker 00

  So I think to the extent that anybody who wants the UN to be effective is the member states. We have to enable the UN to be effective. I'm just, I'm glad to say that so far at least the work we've been doing has been generally well received by the member states. In fact, early this year, the US put forward a resolution which was signed on to by most other countries and unanimously endorsed by the UN member states, which put some basic frameworks and principles and norms, if you like, about how to think about AI. But it's a very complex topic, obviously, to be done.

• 6:34Aria Finger?

  In terms of the global attitudes, is that because you think that folks in the global south potentially have more to gain? It could propel them into the middle class. It could give them English speaking abilities. It could sort of do all those things, whereas folks in the US might be thinking, I have a pretty good life right now and I'm nervous about job loss and I actually don't want to be disrupted because things are going okay. Is that the reason for the difference in attitude or would you say it's something else?

• 6:59Speaker 00

  I think that's part of it. I think other parts of it have to do with the fact that in many places, say in Africa or like in America, just in Brazil recently, often there's a lack of resources, lack of access to doctors, you know, eternal health, you know, any number of challenges, books in schools and so forth. And by the way, the views that I hear from those member states or countries in global south in some ways are not so different than some of the, you might, what you might call left behind communities here in the US, actually. Because when you talk to any of the poor and less affluent communities here in the States, you hear some of the same things.

• 7:43Speaker 00

  So I think the question is, how do you make sure societies and communities anywhere here or anywhere else in the world do have access to resources, help when they need it, whether it's in healthcare, in education, access to information and also accessing their own languages and dialects? That's really, really important.

• 8:04Reid Hoffman?

  And what do you think some of the key learnings have been? Because part of the reason I'm asking this question is you and I talked about this bunch because I rely upon you as the expert to participate in like the UN conversations and kind of thing. What are some of the principles that we should think about as technology inventors, entrepreneurs, that are part of the inclusion for the interests of the global South and the world?

• 8:31Speaker 00

  Well, I think a few things. Part of it is giving communities everywhere the ability to participate in these opportunities. I think those come in two parts. One are the access to, if you like, the ingredients for AI, access to compute, access to models, being able to get access to data that they need to tailor and customize things in their context. And on the access to models, I think that's also the reason why there's an interesting kind of tension around this question of open source. But I think open source is one of the ways to give access to many more people to develop, participate, and be part of it. So that's kind of one of the things that are directly about AI.

• 9:13Speaker 00

  Then you've also got a second part, which are some basic infrastructure needs. I think you and I read in ARI, we can sometimes forget the fact that not everybody has access to broadband, actually, in some parts of the world. And in fact, when I listen to some of the members on the high-level body, they'll even add basic electricity to that and will say, you know, there are some places where without basic broadband and connectivity and infrastructure, now all of this is mostly academic. So I think all of this is very important for inclusion and participation. And then you've got other things like recognizing access to languages and the rich cultural diversity of data and contexts that come from any of these extraordinary places around

• 9:58Speaker 00

  the world. I think all of that is very, very important.

• 10:01Reid Hoffman?

  Yeah, no, it's important for people to know, and you and I talk about it all the time in a number of contexts. So so now for a quick AI experiment. James is co-chair of the UN's AI Advisory Body. You were involved with the publication of a 2023 report entitled Governing AI for Humanity. We ran this paper through one of Google's very cool new AI tools, the virtual research assistant Notebook LM, and asked for a summary specifically tailored for an entrepreneur. Notebook LM turned around bullets of relevant opportunities and challenges for entrepreneurs. And here is one of each.

• 10:41Speaker 03

  First, collaboration and open source. The report underscores the potential of open source AI models and international collaboration to level the playing field. Entrepreneurs can leverage these resources to access valuable tools, data sets, and expertise, fostering innovation and accelerating their ventures. And also, competition and access to resources. Despite opportunities for collaboration, the report acknowledges the competitive nature of the AI landscape with large technology companies holding significant resources. Entrepreneurs may face challenges accessing crucial data, talent, and computing infrastructure.

• 11:22Speaker 00

  Well, I think that that sounded pretty good. It sounded pretty good. I think perhaps things that I may have added to that from an entrepreneur's standpoint, having spent some time talking with entrepreneurs in Latin America and in Africa especially, I think I would also add the need for more robust AI ecosystems, including funding, access to resources, and investment. I mean, these are things you and I have discussed in the past and need to talk about all the time. So the more we can empower entrepreneurs in those places to be able to build companies, participate, and quite frankly, access much larger markets and opportunities. I think that that's something I would add.

• 12:07Reid Hoffman?

  So before we move on to other areas of AI, say a little bit about your specific uses of Notebook. Because this is actually one of the things that I think we try to do with the AI thing is people go, well, I'd say, cool, I can create a sonnet. And they don't realize like today, tangibly, in the stuff you're doing, you can make this really interesting. And here you are, you know, uh, you're running, you know, important AI stuff at Google and you're using it. So say a little bit about how.

• 12:34Speaker 00

  Oh, I use a lot. First of all, let me describe what it does. So it's kind of based on the Gemini Pro model. So it's a, you know, it's very capable model underneath it. It also happens to have this multimodal capability and it has this incredibly long context. And by having a long context, that's what gives it the ability to be able to do the virtual research assistancy thing. So I can actually put, you know, create a notebook that has all my notes around a particular topic. I can add up to 50 sources. Each one can have up to 500,000 words. So if we think about all of that together, that's like 25 million words and that's all my stuff.

• 13:18Speaker 00

  So now it's grounded in my stuff. So now I can have a dialogue with it. So I'll tell you what, at least two or three examples I'm using it for right now. I have a notebook on AI and science where I've put in all the papers and stuff that I'm trying to learn about in science. I'm not as good a biologist. I do a bit of biology, so me read. But so I'm trying to learn as much as I can about biology because of the incredible things that AI is trying to do to biology. So when I put all that in into my AI and science notebook, I can now have a conversations inquiry. I can ask questions. I can ask for connections. I can draft things. And by the way, it has a particularly cool feature, which is I can cite my sources.

• 14:04Speaker 00

  So the nice thing about doing things in a long context, whatever it gives me, you probably saw this in what you did with the UN report. You probably had citations. So whenever it summarizes, if you click on one of those citations, it will actually take you to the sources and the words in the materials you gave it. So it actually cites them. So you don't have hallucinations or any of that. So it's an incredibly useful tool. In fact, there are actually quite a few people using it to do extraordinary things. Walter Isaacson, for example, I think he treated this a while back. He's using it for the work and research he's doing analyzing Marie Curie's journals.

• 14:47Speaker 00

  Stephen Johnson, you know, he got a big data dump of all the Apollo mission transcripts from NASA and he threw all of them in. And basically he's now having a dialogue with all of that for the research and book that he's trying to do. It's very useful. It's very cool.

• 15:03Aria Finger?

  I love that so much. And I think the lack of hallucinations and the fact that you can cite sources, I mean, is a total game changer for folks who are using it in their work context. So we've talked about ways that you are using AI personally. We've talked about some of the limitations, especially in the global south or under-resourced communities. Are there other sort of positive impacts that you're seeing right now from a global perspective from AI? Like the show is possible, we're going to get into the future. But what are you seeing right now about ways that AI is actually helping globally?

• 15:34Speaker 00

  Oh, I can think of several things, Ari. So let me list a few. First, languages. One of the things that's extraordinary, first of all, people have been using Google Translate for a very long time. In fact, up until about maybe two years ago, Google Translate was doing like maybe what, 38 languages? And now it's about two hundred and seventy six. And if you think about the world as a whole, uh, you know, if you want to get to, I think, something like ninety six percent of all the languages spoken in the world, you're gonna have to get to seven thousand languages. So we actually have a moonshot to get to at least a thousand languages pretty soon. So I mentioned that because all of a sudden, people in very different places around the world are able to converse.

• 16:19Speaker 00

  I'll give you another example. As the world started to face the effects of climate change, we know we have many, many more extreme weather events. One way that actually members of my team in Google Research, what began as a small pilot about a year and a half ago, using AI systems to predict floods, what are called riverine floods. We started this as a pilot in Bangladesh. Uh, it worked, uh, and worked in the sense that you can give people up to, uh, at least five days advance notice, which makes a huge difference. And so we went from that pilot in Bangladesh to where today, we're now doing this in over eighty countries,

• 17:00Speaker 00

  giving alerts in areas that are that cover something like seven hundred and sixty million people. This is extraordinary. So, you know, it actually is a way to help. And by the way, of those people, something like, I think, we're we're covering something like eight hundred locations here in the US. It's not just in the global south, but also in the US. There already are incredible ways that this is already making a difference. They don't have to wait for the future, uh, but it's already happening right now.

• 17:26Aria Finger?

  I have to admit, I love that and I just want more people to know about it. It sounds trivial, but honestly, I'm thinking about the global blockbuster twister from this past summer, where they were trying to have scientists, uh, predict when, um, twisters were coming so they'd give people a heads up. And to me, this flood thing is the same thing, except it's it's not science fiction, it's science fact. Like, we're actually doing the things that the global blockbusters are about. People just don't know about it. And so I we just, we need people to know.

• 17:57Speaker 00

  No, we we need to talk about it more. In fact, this summer, actually a month ago, I mentioned wildfire boundaries. We know we have many more wildfires. In fact, we've now gone from up until a month ago, we're covering about fifteen countries. It's now twenty two countries, both in the US and also in Europe, but also in the global south, where we can give people information about wildfire boundaries.

• 18:23Speaker 00

  People may not fully get when they get the benefits of that, that, oh, AI did that. They just go, I've got a flood of earth, or, hey, you know, I've got better treatment. But I think it's important for people to know that there's a lot of this is being enabled by AI.

• 18:38Reid Hoffman?

  You know, one of the things I think you've commented on before is kind of capitalism's challenges in two particular areas. One is kind of economic well-being for all, another one, climate change. So, you know, how do you see the tech community trying to engage AI and science to help mitigate these failures? What's possible? How do we get the tech community more on target with that? How do we get capitalism better aligned, the relationship between AI and capitalism? And how should we be steering that?

• 19:06Speaker 00

  Yeah, I think that's a very important question, Reid. I mean, I'd start with thinking about the quest of AI and the economy. I think, first of all, AI has an extraordinary potential to drive productivity growth, which drives economic growth, which is how societies and communities prosper, basically. In fact, by most estimates, the economic growth from productivity potential is enormous. But it's also important to say those gains won't happen automatically. And this is where what entrepreneurs, companies, and in fact, even what investors have to think about. And some of the ways to get those gains to happen include encouraging wide adoption,

• 19:51Speaker 00

  especially in some of the larger sectors in the economy. It's not enough to, for the tech sector by itself to be using these tools. It matters that we get to sectors like healthcare, retail, and others. So wide adoption is important. It's also important that we're building and focusing on applications that are actually productivity enhancing. And I think this is important for entrepreneurs. To the extent that we're enabling workers and people working very high productivity potential use cases, that becomes very important. And I think you're making sure that, you know, we're building worker resilience and preparedness is also important. Keep in mind that it's going to take a whole lot of other complementary investments for all of this to happen. So I think all of those are things to think about.

• 20:36Speaker 00

  I would say, though, Reid, and this is something you and I talk about all the time. Just because even if these economic gains happen, there's no law of economics or guarantee that will say therefore everybody will benefit. That's going to take work by all of us to make sure that even when the gains happen, every benefit. So part of that is going to include things like making sure everybody can participate, making sure that all communities are part of this, making sure that we're not leaving communities and places behind. And also making sure that workers themselves are ready with skills and capability to take advantage of this.

• 21:19Speaker 00

  We talked about Stanford, our friends, Eric Nelson and others like David Otter, have been doing some very useful and important research on this. And what I like about some of that research, as you know, is that it may be that for the first time, we have a technology that does something different here than what previous digital technologies did. You know, the economists listening to this will know of skill bias, technical change, which is what older technologies did. We showed that historically tended to be the more skilled expert workers were able to take advantage of technology. However, this time around, it may actually be flipped. I mean, David Otter has done some extraordinary work along with Eric, it shows that

• 22:04Speaker 00

  it's now it looks like the less skilled workers are the ones who actually benefit the most from these technologies. David actually wrote a wonderful paper where he suggested that this could actually lead to rebuilding of the American middle class, which is really exciting. So I think we were the precipice here of some exciting possibilities. But you know, as you know, Reed and Arya, it's going to take some smart policymaking. It's going to take some smart actions by entrepreneurs, companies. All of that is going to happen, is going to have to happen together for this to turn out well.

• 22:42Reid Hoffman?

  Yeah, one of the ways to kind of make that last point that you just, you know, the David Otter and point like tangible is if you say, well, if you kind of look at what the generative AI systems are doing today, they create a great assistant at call it ninety percent plus of, you know, kind of percentile of human capabilities. You could have a person plus machine, still be much better. But also, when it gets that skill gap of lower skilled or higher skilled, you go, who is benefited by a amazing assistant that does all of the basics, you know, very well? The answer is the entry skills and that that creates a great kind of leveling, you know, for how this how this operates. And that's one of, I think, the the points of optimism.

• 23:27Speaker 00

  Absolutely, absolutely. And by the way, a lot of this, which is describing a redesign, if you're just theoretical, the early empirical evidence, the few empirical studies actually show exactly what you're describing. Uh, but in fact, I mean, we see a significant level where a lot of programmers now work alongside uh, these AI systems and actually makes them better. Uh, they can do more. So I think all of that is quite, quite exciting. But I think it's also going to be important to to realize that it's important that we have smart policymaking here, uh, and and investments are to guide this, uh, to get to the right place. You know, how do we get this to go right? Uh, I think it's very, very important.

• 24:10Reid Hoffman?

  Let me go to one particular other place of frequent misunderstanding in the press, which is, you know, because people go, okay, power data centers, you know, is this going to be one more nail in the coffin of climate change? I don't understand how AI systems, you know, actually can work very positively on this. So one, you know, how much substance is there to the AI as a negative climate changing? And then two, what can AI be doing to help with climate change? To be helped with greenhouse emissions, et cetera?

• 24:42Speaker 00

  Yeah, I think this one, this one's very important. First of all, data centers in the world make up something like about 1% of all the electricity used in the world. I think it's important to keep that in context. That's 1%. It's what it comes from data center usage of electricity. And of that data center usage, something like about ten percent is actually AI. So I think it's just important to understand that first of all. But but at the same time, it's it's worth acknowledging that, in fact, a lot of the AI systems we're developing are very compute intensive and therefore energy intensive. And in fact, the rate of growth of that compute intensive energy intensive is actually quite high.

• 25:27Speaker 00

  So that's why it is important to acknowledge that. But and also highlight the need to come up with better, more efficient systems, better train, better efficient training algorithms, more efficient architectures and so on. And I think we're all working very hard on that. In fact, you're going to see that start to bend, those curves start to bend pretty soon. Having said that, it's also important to think of the other side. I think by many estimates, including, I think, there was a recent study by BCG, also the Salton Group, which actually highlighted that AI itself can mitigate and help address, make, you know, at least at least a ten percent difference to carbon emissions through any number of ways.

• 26:11Speaker 00

  And I can list a bunch of ways that we can do that. For example, look at climate modeling. There's been some extraordinary breakthrough research by us and many others that's appeared in scientific journals about how AI systems are improving our ability to model climate change. Then you look at mitigation itself. There's an enormous number of things we are doing to make data centers more efficient. There's stuff we're doing, for example, in traffic patterns and drivings. You you wouldn't imagine that the stop-start traffic that happens in cities contributes enormously to greenhouse gas emissions. So we have this project called Greenlight, which is now live in, I think, eighteen cities now,

• 26:52Speaker 00

  where traffic lights and patterns are supported by AI systems to reduce the stop-start traffic. It's a project called Project Greenlight, which again reduces emissions by more than ten percent. So I could go on, I'm sure you've got examples to read, but I think. It's important for people to realize that yes, of course, we should reduce how compute and energy intensive AI itself is, but it's having an outsized effect on addressing and mitigating the effects of climate change. So if you keep those two things in balance, you suddenly realize that the net benefit far, far, far exceeds the energies from AI. That's not to say we shouldn't address energies from AI. We should.

• 27:44Aria Finger?

  I also love that, you know, if you're deciding what to do with your life and you care about climate, like so many young people do, in the past, maybe you studied political science and became an activist, or you went to law school so that you could fight against, you know, oil and gas companies or file, you know, EPA claims. And I think it's really important to tell people, hey, if you want to fight climate change, you could become an AI researcher, you could go into technology. Like, these are ways that we can hugely affect the outcomes of this, and they don't have to pit it against each other. It's not sort of AI versus climate. It's like, this is actually the way that we're going to solve some of these challenges.

• 28:22Speaker 00

  I love that, Ari. One of the things that's exciting to me, and I think to many of us, is the fact that the intersection of AI with so many things is everywhere. AI and climate, whether it's climate change, climate science, climate modeling, AI and biology, AI and physics, AI and quantum. And in fact, it's, you know, many, many aspects and disciplines that people are very excited about now have this extraordinary intersection with AI. If you go to any biology research institute, I have a tweet about it, the Broad Institute of MIT and Harvard, much of the work there now is computational biology. And in fact, you find that people, you know, close to at least a third or more of the researchers

• 29:04Speaker 00

  there are actually computational biologists making use of these techniques. You go to, you find the same thing in climate science now. So, so I think this is true in many, many, many, many disciplines.

• 29:17Aria Finger?

  And so to zoom out a little bit, we've been talking about technological advancements and they should and can be good for the progress of society. We have to get policy involved. We have to make good decisions. We have to have government work with businesses. But also, especially with AI, because as we said, it's an intelligence, it stimulates these larger political and philosophical questions. How has working in and around AI for so long changed how you think about humanity? Or what have you learned about humanity in the context of all these AI developments?

• 29:53Speaker 00

  That's an important question. I think there are questions about risks and complexities which you should discuss, but there are also some quite deep philosophical questions. And those to me include the questions about the ethics of all of this and the societal implications of all of this. But even the question of, I think AI in some ways is putting a mirror in front of us, allowing us to look at ourselves as humanity. Questions about what does it mean to be intelligent? What does it mean to be creative? What does that mean to be empathetic and thoughtful? Because in many ways, AI is having us reframe many

• 30:34Speaker 00

  of those questions. One of the things that's been fascinating, Aria, to me is if you look at the history of AI, which has been quite documented by many, there's a book, by the way, Kamala McCordy's book is a classic. I'd encourage you to read it. It's called The Machines Who Think. She kind of chronicles the history of the field. But one of the things you get from that is how for the last 50 years, as AI has made some progress, greater or less, it's courses to reframe what it means by creativity. So there was a time when we used to think creative men being able to do these kinds of things or being intelligent men doing these kinds of things. Then you suddenly find that AI is now able to do some of those things. And then you kind of re-shift and reframe the

• 31:18Speaker 00

  question. All of this is giving us ways to rethink it, look at ourselves, and rethink through these questions. Even take these issues of bias and fairness. It's easy for us to say we want systems that are fair or not biased. Okay, what does that mean? The humanity has been trying to figure that out for thousands of years. So I think there's some deeper questions that we're having to grapple with here. And I think it's a good thing. I call it point for reflection. To go back to the UN thing, it's actually been quite useful for me, by the way, hearing from so many people from around the world who don't sit where I sit, who don't do things that I do, who don't see the world the way I do,

• 32:02Speaker 00

  to hear both about their excitement, but also concerns about this technology and realizing that there are ways for us to work together to get this right.

• 32:14Aria Finger?

  So are there specific things that you think we should be doing more across the industry to mitigate risks? Like, we all agree there's enormous potential and talking to people from 33 countries on your advisory panel is critical. Like, what we should be specifically doing to ensure there's no bad actors, to ensure that we get everyone's point of view, just to create this best possible future?

• 32:37Speaker 00

  I can think of a number of things, Aria. I think part of that is agreeing as a community and an ecosystem on standards, safety mechanisms, principles for evaluating the systems we're developing, protocols for when things go wrong, finding ways that are appropriate to share information and experiences. I think it's also important to agree on things like, you know, when is it appropriate to do what? Can we all agree, for example, you know, as the technology develops, if we can do things like watermarking, if we can do things like provenance and share information about, about all of that, we should do that. I also would say that in a different direction,

• 33:22Speaker 00

  I think it's going to be quite important to make sure we are necessarily going to making sure that these systems and capabilities benefit people who don't normally participate or who don't normally benefit from this. So I often want to tell my colleagues that just to assume that Commercial interest alone will get to everybody. I don't think that's the case. I think there are some applications, some uses that should be considered in the public interest that, you know, simply commercial approaches won't get to, right? You know, if, yeah, I'll use my language as example, if, you know, if, if we only

• 34:03Speaker 00

  cared about languages and translating languages that only benefit the places where you can hope monetize, you'd probably stop at about forty, the language number forty. You wouldn't bother to go to a few hundred, right, or even aim to go to a thousand. So I think many more examples of going to those in the public interest uses and use cases, I think is, is another thing that this community can do. I mentioned one other quick thing, Ari. I think more than ever, this is going to become a very interdisciplinary field. I think we have to involve philosophers, ethicists, not just purely technologists. I think these questions have become so multifaceted.

• 34:47Speaker 00

  I don't think any discipline, computer science, has any monopoly on how to think about these things. So I think the more we can involve a diverse set of, you know, people from different areas of expertise, I think is really, really important.

• 35:02Reid Hoffman?

  Yeah, I think one of the things, like, people naturally think what this kind of AI revolution is, is a further kind of quadrupling down of how important CS is and machine learning and everything else. And of course, those are very important. What is almost uncommoned on, what you're gesturing at, is actually the importance of a variety of the humanist disciplines. We need to figure out how to get, you know, the question about the human search for truth is allowing a discourse between positions and allow these, you know, like, people to have that discourse is what's most important. And that's what we're going to be needing to build into these AI tutors.

• 35:45Reid Hoffman?

  Like, all of these humanist disciplines are now even more important in the age of AI.

• 35:50Speaker 00

  Oh, oh, and you know, absolutely, absolutely. And I think it's also one of the other thing that's also going to be collectively important. Making sure that the questions we're exploring and try to understand keep up with where the keep up with is that the technology are. Sometimes there's a sense that the questions we're grappling with, whether it's ethically or philosophically, have to do with the outputs of these systems. Sure, that's true, right? In terms of the content and all the rest of it. But that question's already just one of several other questions. As these systems become more agentic, for example, there are going to be questions about what actions are appropriate. So I think the questions keep evolving and shifting as the technology advances.

• 36:32Speaker 00

  So I hope that we can find ways to, you know, also be in dialogue with philosophers and policymakers and others as the technology advances so that we're not solving for technology questions from 10 years ago, but actually solving for the ones today.

• 36:50Aria Finger?

  And so internally at Google, how do you grapple with these questions? And then also like what's to come? Like, how, how is, you know, Notebook LM going to be rolled out globally? What is the future of that going to be for Google in particular?

• 37:04Speaker 00

  Oh, and the case of Notebook LM, actually, it's pretty cool. So it's now available in 200 countries already, which is very good. You know, it's going to have this audio capability. So you can actually get audio overviews that create a conversation, you know, discussing whatever's in your sources with AI-generated voices, which is actually also very cool. But I think to your broader question about how are you thinking about this at Google? I think, you know, as I mentioned earlier, we're trying to do two things at the same time. On the one end, we are trying to be as ambitious as we can and as bold as we can about can we advance all the ways that we just can benefit people and society.

• 37:48Speaker 00

  So in that regard, we think about. Benefiting people, I mentioned, you know, trying to get to a thousand languages. We're trying to help, you know, people with disabilities to use this technology or to not to be left out. So there's all these things for people. Then there are all these things to power economies from, you know, small businesses to large companies that hold economies. The economy is important. And then all the science things we're trying to do, you know, from alpha four to material to climate, et cetera. And all these societal things. So we're trying to be bold and ambitious in that way. At the same time, uh, we're trying to be responsible. And this is, you know, hard work. We're trying to, you know, improve the performance of these systems, uh, get it right. We don't always get it right.

• 38:31Speaker 00

  Uh, make sure that, you know, there's accurate, they're not biased, uh, they don't give you weird results, which they have, you know, on occasion. And trying to work and fix that. But we're also trying to make sure we're responsible in how we deploy these systems, uh, to avoid misuse. So we've been working on things like synth ID, for example, to watermark outputs of images and so forth. So we're trying to work on that. And then on some of these societal complexities, we're trying to also help with scaling and workforce capability and make sure workforces are resilient. So, so we've got this kind of two part set of things we're trying to do.

• 39:11Speaker 00

  Uh, the thing about those, that those kind of bold things, the responsible things are, is we don't see those as intention. We actually like the constructive tension that that creates, actually. Uh, and we want to embrace both things at the same time. Doesn't mean we get it right all the time, but that's, that's kind of our North Star.

• 39:31Reid Hoffman?

  How do we think kind of epistemology, learning and truth are going to evolve through the use of our modern AI tools? And, you know, what's the way that people should understand it? What's the way that we should shape it? It's obviously pretty core to, you know, the Google mission, but it's also, of course, super important to humanity. So a few remarks there and then we'll get to rapid fire.

• 39:57Speaker 00

  There has to be a way to truthfully and factually understand the real world and the world as it is. And so that that's got to be how, I mean, that's how society, humanity has progressed. I mean, there's got to be a shared sense of truth and objectivity that reflects the real world. That's why I think being focused on factual, truthful, science-based perspectives, I think it's got to be a north star for what we think about this, even in the age of AI. You know, I'm always amused to read by, I think, one of the, I remember getting a critique from somebody who said, one of our AI systems was biased because it said climate change is real, right? And to think, okay, right.

• 40:40Speaker 00

  But I think those are much more normative societal questions. I, I just hope we keep to the idea of testable science. Those kinds of foundational principles are quite important.

• 40:58Reid Hoffman?

  All right. So rapid fire. Is there a movie, song, or book that fills you with optimism for the future?

• 41:06Speaker 00

  Uh, well, movies, films, the Star Trek. I mean, Star Trek always makes you very optimistic, just because it's, it's got technology in it. It's optimistic. It's thoughtful. It's complex. It's complex. I always, I always feel, I like that.

• 41:18Aria Finger?

  And so is, what is a question that you wish people would ask you more often?

• 41:23Speaker 00

  What could go right? I wish people could ask that more. What could go right? Uh, a second question, I'd say, I think it's important to ask, people to ask me and us, what are some of the exciting things in AI outside of chatbots? I think somehow we found ourselves into this corner where we think all of AI is chatbots. Uh, it's a much wider field and there's so much more going on. I wish people could ask more about that.

• 41:51Reid Hoffman?

  Yes. Well, as you noted, we didn't ask you one question about chatbots on the entire podcast.

• 41:56Speaker 00

  Thank you.

• 41:59Reid Hoffman?

  You know, I know you can answer this question in many different ways. So I'm curious which way you'll do it. But where do you see progress or momentum outside of your industry that inspires you?

• 42:08Speaker 00

  In science. I think there's an exciting time for science, actually. AI is an even part of that, but there's certainly an exciting scientific possibilities ahead of us.

• 42:20Aria Finger?

  I mean, this whole podcast, I'm like, I gotta go back and I gotta get a degree. I mean, the science is so fascinating. And so for our final question, can you leave us with a final thought on what you think is possible to achieve if everything breaks humanity's way in the next 15 years? And what's the, what's the first step in that direction?

• 42:39Speaker 00

  I think all the fine ways to include everybody to participate in this exciting thing, and I think access to language, access to opportunities, I think will revitalize growth of the middle class. I think David Autor is perfectly correct, although we have a chance here to revitalize the growth of the middle class. I also think that science is going to make, and the impact of science in society is going to be pretty profound. I think all of these things could really break well for humanity. I'm very excited about all of that.

• 43:12Aria Finger?

  I love it. James, thank you so much. We really appreciate it.

• 43:16Speaker 00

  Well, thank you both.

• 43:17Reid Hoffman?

  I look forward to seeing you again soon. Possible is produced by Wonder Media Network. It's hosted by Aria Finger and me, Reid Hoffman. Our showrunner is Sean Young. Possible is produced by Katie Sanders, Edie Aller, Sarah Schlied, Adrian Bain, and Paloma Moreno Jimenez. Jenny Kaplan is our executive producer and editor.

• 43:44Aria Finger?

  Special thanks to Asuria Yalamanchili, Sayida Sebibeva, Ian Ellis, Greg Beato, Ben Relis, Parth Patel, and a big thanks to Carrie Wang, Rochelle Lacroy, Carrie McHugh, Sarah Portic, and Little Monster Media Company.