TED Radio Hour - What one founder's past says about AI's future

Episode Date: December 20, 2024

To understand AI, you need to hear Mustafa Suleyman's story. From youth hotline founder to DeepMind creator and Microsoft AI chief, his journey explains AI's evolution and his plans for its future.TED Radio Hour+ subscribers now get access to bonus episodes, with more ideas from TED speakers and a behind the scenes look with our producers. A Plus subscription also lets you listen to regular episodes (like this one!) without sponsors. Sign-up at: plus.npr.org/ted.Correction: This story incorrectly states that DeepMind was one of the first AI companies. DeepMind was not one of the first AI companies, it was one of the first AGI—artificial general intelligence—companies.See pcm.adswizz.com for information about our collection and use of personal data for sponsorship and to manage your podcast sponsorship preferences.NPR Privacy Policy

This is the TED Radio Hour. Each week, groundbreaking TED Talks. Our job now is to dream big. Delivered at TED conferences. To bring about the future we want to see. Around the world. To understand who we are. From those talks, we bring you speakers and ideas that will surprise you.

You just don't know what you're going to find. Challenge you. We truly have to ask ourselves, like, why is it noteworthy? And even change you. I literally feel like I'm a different person. Yes. Do you feel that way? Ideas worth spreading.

From TED and NPR. I'm Manus Zameroidi. It was about a year after 9-11. This is Mustafa Suleiman. In 2002, Mustafa was a student at Oxford University when a friend decided to start a helpline for young British Muslims. A good friend of mine at the time had just started the first prototype one evening session.

I think it was a Thursday evening of offering counseling services on the phone. Mustafa was intrigued. He ended up joining the project to help grow it. And we became essentially co-founders. Quickly, their service was in hot demand. It was a pretty remarkable time because young British Muslims were feeling judged by 9-11 as though they were responsible or somehow complicit. There was a lot of Islamophobia.

But what they found these young people needed was really just someone to talk to. Most of the challenges that they were working through were the sort of conventional things that a young person might deal with, like relationships, bullying, self-harm, increasing homophobia in the community. And fundamentally, I think people were just looking to find support in a language that made sense to them, you know, like a London accent using sort of Islamic terminology and having a broader understanding of the, you know,

whether it was a Bangladeshi or Pakistani culture or an Arab culture, combined with the religious foundation, and just being able to reflect back that empathy went a long way. 20 years later, the Muslim youth helpline is still going strong. and Mustafa Suleiman now lives in Silicon Valley, where he's been working on another kind of helper. I'm the CEO of Microsoft AI, and I work on building new AI systems and products. These days I'm working on an AI companion. With artificial intelligence and the backing of one of the biggest tech companies in the world, Mustafa Suleiman wants us each to have our own digital assistant.

I can get philosophical, creative, or poetic, if you like. Available 24-7, kind of like a personal helpline. I can be your sounding board as you work through challenges. I can write stories, brainstorm ideas. Think of me as your superpower in your pocket. How can it serve you? How can it support you? How can it save you time?

How can it reduce your anxiety? How can it help make you feel smarter and more confident and more prepared to go about it? your day with no judgment, no pressure. Hey, Manusch, nice to hear from you. What's on your mind today? Microsoft calls this assistant co-pilot, and it was released at the end of 2023. I've been playing around with it. Well, I need to interview Mustafa Suleiman, your creator.

What do you think I should ask him? I go for topics like his vision for AI's role in society, the balance between innovation in ethics and what he sees as the most exciting future possibilities for AI. Okay. These are good. Now, can you please write the intro for this episode? Sure thing. Here's a possible intro.

Hey, it's Manus Samaroti, and today's TED Radio Hour is dedicated to a deep dive with none other than Mustafa Suleiman, co-founder of Deep Mind and author of The Coming Wave, Mustafa has been a pivotal force in AI, driving both groundbreaking advancements and sparking intense debates. In this episode, we'll discuss the promises and perils of artificial intelligence and whether we're truly prepared for the wave of change it's bringing. Is AI evolving too fast for society to adapt? Are we doing enough to mitigate the risks? It's an hour of insight, reflection, and some tough questions. So, let's jump in. Okay, that was pretty good. As co-pilot noted, Mustafa Suleiman is a long-time pioneer in the field of AI.

After he dropped out of Oxford University, he worked for London's mayor. He was even part of an NGO's delegation to the Copenhagen climate negotiations in 2009. He wasn't impressed. There were so many sort of difficult egos and entrenched sort of interests. and I just felt a sense of sort of frustration and I felt kind of demoralized after that experience. Mustafa wanted to change the world. Governments and non-profits didn't seem terribly effective.

But there was one company that was managing to change the behavior of millions of people. Facebook was exploding at that time. It had got to 100 million monthly active users. in the course of a couple of years. And it was pretty clear to me that that was going to have a profound impact, more so than anything I could do in sort of the world of social activism. Around then, Mustafa started hanging out with the older brother of a friend, a software genius named Demis Hasabas,

who had been designing computer games since he was a teen. Demis saw how to make Mustafa's vision possible. And he was coming at the world from a completely different perspective, which I found really interesting. His belief was that we would just simulate, you know, complexity in the world. This new suite of methods in AI self-learning systems were coming up strong and looking likely to work. And, you know, we really just clicked on that kind of technical

and socio-cultural view of how to positively impact the world. Together, the two decided to start one of the first AI companies ever. They called it Deep Mind. It was very clear from that, even those early stages, that if we got the technology right, and it was going to be this decade that led to major, major breakthroughs, then the consequences for humanity were going to be significant. I've been lucky enough to be working on AI for almost 15 years now.

Mustafa Suleiman picks up the story from the TED stage. Back when I started, to describe it as fringe would be an understatement. Researchers would say, no, no, we're only working on machine learning because working on AI was seen as way too out there. In 2010, just the very mention of the phrase AGI, artificial general intelligence, would get you some seriously strange looks. and even a cold shoulder, you're actually building AGI, people would say, isn't that something out of science fiction?

People thought it was 50 years away or 100 years away, if it was even possible at all. Talk of AI was, I guess, kind of embarrassing. People generally thought we were weird. And I guess in some ways, we kind of were. The ironic thing is that many people still don't think it's possible, many people still think that we're crazy. And at the time, people really thought we were crazy. I mean, it was so far out there.

It was really strange. And we were a strange group of people. I mean, you know, Demis, our third co-founder, Shane, who basically a mathematician spent his entire career thinking about how to formalize a definition of intelligence that we could use to measure our process. progress in the company. We were misfits and kind of outsiders and there weren't very many people willing to back us. So naming the company Deep Mind, what was it that you pictured in your mind that

you hoped to achieve? Yeah, I mean, we wrote our business plan in the summer of 2010 and took it to Silicon Valley to shop it around a bunch of people. And the strap line for the business plan was, you know, building artificial general intelligence safely and ethically. And then that evolved into a two-line mission, which was solve intelligence and use it to make the world a better place. And that dual frame was kind of the foundation of the company. That, you know, our belief that science and technology was the engine of progress. And there are some downsides, but I certainly think this is the – this is the engine of creating civilization in a more healthy and sustainable way for the very long term.

And, you know, if you think about it, everything that is of value in our world today is a product of us humans being able to take materials from our environment and recompose those into products and services and other compounds that are useful to us, from our lights, dark, carpets to everything that you see in front of you today. In 2014, DeepMind ended up being acquired by Google. That must have been huge for you in terms of money, resources. You were off to the races. Yeah, it was a huge event. I mean, it was the largest acquisition Google's ever made outside of the U.S. We became Google's primary AGI bet.

and we were empowered with tremendous resources, both people and computation, to go and both do the hard science, but also work on really important applied practical product problems. And that's where I kind of really honed my craft, if you like. You know, as a product maker, it was just the most amazing experience. And as early as 2015, I actually ran a, a hackathon project in my applied group at DeepMind.

And the theme of our hackathon was to find high-impact ways of using our technologies for good. And so there were lots of application. There were sort of lots of prototype hackathon experiments in healthcare or in energy systems, both of which went on to become significant parts of DeepMind applied. And, you know, my group published three, papers in nature showing human level performance on, for example, classifying eye diseases, the top 50 eye diseases from OCT scans, three-dimensional eye scans, showing that we could perform as well as a panel of radiologists in identifying cancerous tissues in mammograms,

showing that we could predict the onset of sepsis and acute kidney injury. as well as the top, you know, renal doctors using vast amounts of data. And this was way back in sort of 2016, 2017, 2018, and really helped to kind of lay a foundation for the application of large-scale machine learning to, you know, tough social problems. And that was very much my motivation. In a minute, more of the incredible breakthroughs that Mustafa and DeepMind had at Google that changed the way medicine works today

and why Mustafa ended up leaving the company. The company was just frankly being too slow to get things into production. Today on the show, the CEO of Microsoft AI, Mustafa Suleiman, and the future of artificial intelligence. I'm Manus Shumeroody, and you're listening to the TED Radio Hour from NPR. Stay with us. Hey, it's Manus. This past year, we have brought you stories about AI, relationships,

climate change, neuro technology, dinosaurs, privacy, human behavior, and even one about what it means to create thriving public spaces. All of these public spaces that we take for granted, all of this social infrastructure.

We work really hard to bring you all of these stories because that is our public service, kind of like a park or a public library. And libraries are these beautiful hubs that can take on the of whatever community that uses them really needs.

Public media is infrastructure that we all can use. It's free and it's for everyone. And our mission is to help create a more informed public. So this time of year, I want to thank you for listening and to say that one of the best ways you can support this public service is by signing up for NPR Plus. With one subscription, you are supporting all the work that NPR does and you get sponsor-free episodes and perks across NPR's catalog of 25 podcasts.

It's a great time to join because throughout December and January, plus listeners have access to even more insights from TED speakers to help you kick off 2025 right, from making big life decisions to being more hopeful, to carving out time for what is important to you. We have got you covered. Just visit plus.npr.org. You can also find the link in our episode notes. And the other way you can give is to make a donation at donate.npr.org.

Your gifts are tax deductible either way. Thank you so much for being here. And now let's get back to the show. It's the TED Radio Hour from NPR. I'm Manushe Zamorodi. Today on the show, we're spending the hour with Mustafa Suleiman, the CEO of Microsoft A. who was the co-founder of one of the first AI tech companies ever, Deep Mind.

The company was acquired by Google in 2014. And Mustafa and his co-founder, Demas Hasabas, went on to have numerous scientific breakthroughs, including a project called AlphaFold. AlphaFold uses AI to figure out incredibly complex protein structures in molecules, the building blocks of every biological process in our bodies. bodies. It won Hasabas the 2024 Nobel Prize in Chemistry. Here he is on the TED stage. The rule of thumb is that it takes one PhD student, their whole PhD, so four, five years, to uncover one structure. But there are 200 million proteins known to nature. So you could just,

you know, it would just take forever to do that. And so we managed to actually fold using alpha fold in one year, all those 200 million proteins known to science. So that's a lot of it. a billion years of PhD times. So. Alpha Fold was just one of the ways that computers were solving biological problems, says Mustafa Suleiman. I mean,

this really, Alpha Fold is really the first time we started hearing about, I guess, the term computational biology, this idea of using tech and science to rethink how biology works and then getting it out into the world, changing the way we

treat diseases, or maybe be developing crops that are more resilient, all kinds of ways that what you figured out at DeepMind would potentially change the world, which was the goal? Absolutely. That was the core mission of the company from day one. Like, how could we take that process of synthesis and prediction and turn that into a general purpose system that could use all the data and all the compute that we have in the world to try and solve these very hard problems,

whether it's growing crops in drought-resistant environments, whether it's more efficient ways to do water, desalination, whether it's long-term battery storage. But in 2022, Mustafa decided to leave Deep Mind and Google. There were reports that he clashed with Google leadership over using technology for military projects and that some employees had complaints about his management style. Mustafa says he was simply frustrated with the pace of innovation

and was anxious to get an AI product out into the world that everyone could use. You know, I really wanted to scale these large language models and get them into production as quickly as possible. And the company was just frankly being too slow to get things into production. And I felt that it was an opportunity to, sort of have all the resources that, you know, that I needed. So I went out and fundraised over a billion dollars and, you know, just work extremely fast with a very small team. And I think I, at the time, and I am still now, very much a believer that these AI companions are going to become an everyday part of our lives.

Mustafa's AI called Pi debuted in 2023. Chad bots have taken the world by storm. Several months later, rival company OpenAI put out its AI product, chat GPT, for free. Chat GPT answers questions and writes essays. They read like the work of a human. An artificial intelligence tool has the potential to change the way we live. This was the point when AI went mainstream. And Mustafa's small, nimble startup just didn't have the resources of these bigger tech companies.

The truth is the pace of big tech had really accelerated in 2023. Google, Microsoft, OpenAI, everyone was really going for it and essentially made these huge models available to everybody for free, which sort of changed our business model. At that very tough moment, Microsoft CEO, Satya Nadella approached Mustafa and made him an offer. You know, the offer that he made was, well, look, we've got all the computation, all the data and all the distribution that you could dream of, come and run consumer products here at Microsoft and really build the future of AI here.

And that was a huge offer. So the product that we're building at Microsoft is called Copilot. And the reason why we've called it that is because Copilot is an aid. You know, it's a conciliary. It's an assistant. It's in your corner, aligned to your interests, on your team, backing you up. Yeah, I actually talked to co-pilot about my interview with you. How do you feel about Mustafa Suleiman, considering that he is your creator? I'd say I'm intrigued by Mustafa. Just so you know, she said she's intrigued by you, just as she is with many innovators in the tech world.

And his work has pushed the boundaries of what's possible. I will say it was very helpful in terms of suggesting topics to bring up with you. But it could not check the weather. And is that simply because of how the access or the data, the information that the agent or co-pilot has access to, that it's not immediate yet? Yeah, there are loads of weaknesses of our co-pilot today. But all of those things will come. I mean, we're working on all of those things that will have permanent members. session to session infinitely across time. It'll be able to access, you know, your files and folders

if you choose to make those available, your email, your calendar, and be able to browse the web for you in the future. And all of that will be integrated into these like seamless conversational, very friendly, polite experiences. I literally was talking to it last night about what to watch and we were going back and forth on whether I would enjoy Pan's labyrinth and whether I've got the time to watch all of Dune because I haven't seen Dune 1 or Dune 2. I mean, it reminds me a little of the hotline for Muslim youth that you're describing. It's helpful. It's infinitely patient.

It's supportive. Are we talking mostly about companionship and mental health resources that this can provide? Or how do you see it? No, I think that the cool thing about co-pilot is that it doesn't judge you for asking a stupid question. Even if you have to ask that question three times over in five different ways, you know, even your best friend might be like, hey, come on, man. I mean, you're asking me this again, seriously. Whereas, you know, co-pilots here for it. There's obviously some similarities to stuff I've done in the past. And I guess it's kind of inspired by nonviolent communication,

if I'm honest with you. It's certainly not like a mental health app or, you know, anything like that. It's just got a little bit of kindness and empathy. It's got some emotional intelligence. And I think that's no bad thing. Gosh, is that where we've gotten to that technology has to tell us how to communicate with each other better nonviolently? Well, it doesn't tell us. It just demonstrates. Yeah, exactly. It demonstrates. But that's what technology's always done. The choice architecture, the buttons, the colors, the language, that is shaping our behavior, whether it's an infinite scrolling feed or whether it's a encouragement to go and film your lunch for Instagram or create a little video for TikTok. I mean, all of those inputs shape behavior. And so we have to

be super thoughtful about what those inputs actually are because technology shapes us in return. and we're in this constant cyclical sort of feedback loop and interaction. And that is kind of what's propelling us forward as a civilization. And it's very powerful. And so far, so good. It's actually been, you know, very, very productive over the last couple centuries. Science has massively delivered for us. But we shouldn't just assume that that's going to happen naturally or inevitably.

We have to really be deliberate and thoughtful. about what the consequences are ahead of time. In the book that you wrote that came out in 2023, you really tried to put what's happening with AI in a historical context. So if the printing press let people own and share information and the personal computer let people search and disseminate information, tell me how you're thinking you can explain to people what AI will do for people now. Each new wave of technology is fundamentally a new interface. It's a new interlocutor, a translator, a way of sort of accessing and creating new information, new tools, new knowledge. So if the last wave of social media and web search help people to access information, this wave is going to help us to invent and create new ideas, be it in science,

or in culture and media and entertainment. And I think everybody is ultimately going to have an AI companion, just as we have a search engine or a smartphone, and just as we use a browser, you'll just ask your computer in natural language, can you write that contract and check that it's okay? Can you create that new piece of software for me? And you're just going to describe what it is.

Can you help me plan that trip? you know, for my parents that are coming into town. So, you know, that kind of breakthrough is a change in the interface, which changes itself what we can actually get done. And I think it's going to be pretty transformational. With the invention of computers, we quickly jumped from the first mainframes and transistors to today's smartphones and virtual reality headsets.

Information, knowledge, communication, computation. In this revolution, creation has exploded like never before. And now a new wave is upon us, artificial intelligence. These waves of history are clearly speeding up as each one is amplified and accelerated by the last. And if you look back, it's clear that we are in the fastest and most consequential wave ever.

The journeys of humanity and technology are now deeply intertwined. In just 18 months, over a last year, a billion people have used large language models. We've witnessed one landmark event after another. Just a few years ago, people said that AI would never be creative. And yet AI now feels like an endless river of creativity, making poetry and images and music and video that stretch the imagination.

People said it would never be empathetic. And yet today, millions of people enjoy meaningful conversations with AI's talking about their hopes and dreams and helping them work through difficult emotional challenges. AIs can now drive cars, manage energy grids, and even invent new molecules. Just a few years ago, each of these was impossible. And all of this is turbocharged by spiraling exponentials of data and computation. Last year, Inflection 2.5, our last model, used five billion times

times more computation than the DeepMind AI that beat the old school Atari games just over 10 years ago. That's nine orders of magnitude more computation. 10x per year, every year for almost a decade. Over the same time, the size of these models has grown from first tens of millions of parameters to then billions of parameters, and very soon tens of trillions of parameters. If someone did nothing but read 24 hours a day for their entire life, they'd consume 8 billion words. And of course, that's a lot of words.

But today, the most advanced AIs consume more than 8 trillion words in a single month of training. And all of this is set to continue. The long arc of technological history is now in an extraordinary new phase. I think the way to think about this is that we're at the very, earlier stages of the development of this technology. Today, when you need to go and sort of ask a question like that, you go to a search engine like Google, you type in a query, you get these 10 blue links, you have to then go to the webpage, you have to read all of this sort of complicated stuff,

formatted in 25 different ways. And that takes time. And it means that you don't always want to invest two or three minutes to go and make sense of that. You don't always have the energy for it. Whereas you can always just like quickly send off a text or quickly make a phone call or or leave a voice note. So I think it's lowering the barrier to entry to access high-quality information. But how do we know it's high-quality information? How do we make sure of that? Well, there was an amazing study two months ago

that was published in the journal Science, which showed that for over a thousand participants who held strong conspiracy theory beliefs. So this is things like flat-earthers or all kinds of things about the COVID vaccine, etc., etc. And after a month of conversation with an AI that had been primed to talk about these conspiracy theories, it reduced the belief in those conspiracy theories by 20%, which is pretty significant. And I think shows the first glimmers of what it's like to have that kind of patient,

insistent, deliberate access to high-quality information. and in the case of this study, human fact checkers, expert human fact checkers, went back and read over the transcripts. So my point is, the value of these systems is that they actually are more accurate. And that's only going to continue. They're more accurate than your average human are answering any question today. You know, we've sort of passed the Turing test because they're more, you know, fluent and conversational. Dan, we ever thought they were going to be, which is pretty remarkable, and that's only going to continue.

So what does this mean in practice? Well, just as the Internet gave us the browser and the smartphone gave us apps, the cloud-based supercomputer is ushering in a new era of ubiquitous AIs. Everything will soon be represented by a conversational interface, or to put it another way, a personal AI. And these AIs will be infinitely knowledgeable, and soon they'll be factually accurate and reliable.

They'll have near-perfect IQ. They'll also have exceptional EQ. They'll be kind, supportive, empathetic. These elements on their own would be transformational. Just imagine if everybody had a personalized tutor in their pocket and access to low-cost medical advice. A lawyer and a doctor, a business strategist and coach, all in your pocket 24 hours a day.

But things really start to change when they develop what I call AQ, their actions quotient. This is their ability to actually get stuff done in the digital and physical world. And before long, it won't just be people that have AIs. Strangers, it may sound, every organization from small business to non-profit to national government, each will have their own.

Every town, building, an object, will be represented by a unique, interactive persona. And these won't just be mechanistic assistance. There'll be companions, confidants, colleagues, friends and partners as varied and unique as we all are. At this point, AIs will convincingly imitate humans at most tasks. And we'll fill this at the most intimate of scales, an AI organizing a community get-together for an elderly neighbor,

a sympathetic expert helping you make sense of a difficult diagnosis, but we'll also feel it at the largest scales, accelerating scientific discovery, autonomous cars on the roads, drones in the skies, they'll both order the takeout and run the power station. They'll interact with us, and of course, with each other. They'll speak every language,

take in every pattern of sensor data, sites, sounds, streams and streams of information far surpassing what any one of us could consume in a thousand lifetime. When we come back, the remarkable, but also terrifying visions of what AI may be capable of in the future, dystopian scenarios that might make you want to unplug all your devices forever. Are they realistic or just fear-mongering?

The CEO of Microsoft AI, Mustafa Suleiman, weighs in. I'm Manus Shumeroi. And you're listening to The TED Radio Hour from NPR. We'll be back in a minute. It's the TED Radio Hour from NPR. I'm Manoosh Zamorodi. Today on the show, the future of artificial intelligence. We're spending the hour with the CEO of Microsoft AI Mustafa Suleiman.

In 2023, Mustafa wrote a book called The Coming Wave. In The Coming Wave, Suleiman delves into scenarios where AI... I asked Microsoft. co-pilot to help me talk through the reasons why Mustafa says AI could be so dangerous. Imagine a world where artificial intelligence has seamlessly integrated into everyday life. The first is what he calls asymmetry, meaning vast amounts of data that could get into the hands of a single person who could go rogue. Because we're now concentrating knowledge and capability into smaller. and smaller units that are transmissible.

You can download open source models, you know, for a few gigabytes and put it on a thumb drive. And, you know, so that's the kind of asymmetric impact because that therefore means that individual actors or smaller groups of actors can have a massive one-to-many broadcast effect. This is AI unleashing computing power to many, kind of like the anonymous hacker who could bring down an entire electrical grid, that kind of thing. Healthcare systems could be overwhelmed. Economies paralyzed. Is that right?

That's right. You can just sort of self-start immediately. Okay, so number two is hyper-evolution. Yeah, software itself just evolving in hyper-real time. In this scenario, a rogue scientist or a bioterrorist could design a pathogen with specific traits, like heightened transmissibility or increased lethality. Or new iterations of weapons and surveillance could be developed. To track people's movements, conversations, and even emotions through their online activities.

And put to use before authorities have time to test them or put any safeguards in place. A small programming error or a hatch system could result in catastrophic loss of life. How could a country build a defense system if it doesn't even know what it needs to defend its against. Which brings us to the third dangerous attribute of AI, Omni Use. One single kind of technology will be able to do everything. Your AI companion will make phone calls, call other AIs, or call other humans to check on, you know, stock or, you know, sort of availability in a hotel or get some advice from a local tour guide on where you're going to visit. Or it will send emails, or it will, you know, talk to another database or a back end.

All of those things are essentially the AI learning to act on your behalf. Sure, but what if your AI also talks to another database or makes a trade on your behalf and decides to shut down the entire stock market, which relates to the fourth and final, maybe scariest attribute, autonomy, AI that takes action on its own without. a human's go ahead. Autonomous weapon systems operate on battlefields, making split-second decisions about targeting and firing without direct human oversight.

I mean, autonomy is one of the core characteristics of these systems. It's a long road, and it's going to be many years before these systems are truly autonomous, and we want to be very careful about that, because we have to ask ourselves, like, what is the added benefit of the system operating autonomously, and is that worthwhile? Is it safe? Is it stable?

Is it controllable? Can we really trust that autonomy? So, yeah, I think it's one that we'll have to be very careful of. The societal disruption could lead to widespread fear, mistrust, and geopolitical tension. Suleiman and other experts argue for robust oversight and ethical guidelines to ensure that these powerful technologies are used responsibly and safely. For years, we in the AI community have had a tendency to refer to this as just tools, but that doesn't really capture what's actually happening here.

Here's Mustafa Suleiman on the TED stage. AIs are clearly more dynamic, more ambiguous, more integrated, and more emergent than mere tools, which are entirely subject to human control. So to contain this wave, to put human agency at its center, and to mitigate the inevitable unintended consequences that are likely to arise. We should start to think about them as we might a new kind of digital species.

Now, it's just an analogy, it's not a literal description, and it's not perfect. I mean, for a start, they clearly aren't biological in any traditional sense, but just pause for a moment and really think about what they already do. They communicate in our languages. They see what we see.

They consume unimaginably large amounts of information. They have memory. They have personality. They have creativity. They can even reason to some extent and formulate rudimentary plans. They can act autonomously if we allow them. And they do all this at levels of sophistication that is far beyond anything that we've ever known from a mere tool.

And so saying AI is mainly about the math, or the code is like saying we humans are mainly about carbon and water. It's true, but it completely misses the point. And yes, I get it. This is a super arresting thought, but I honestly think this frame helps sharpen our focus on the critical issues. What are the risks? What are the boundaries that we need to impose? What kind of AI do we want to build or allow to be built?

This is a story that's still unfolding. Nothing should be accepted as a given. We all must choose what we create, what AIs we bring into the world, or not. These are the questions for all of us here today, and all of us are live at this moment. You lay out 10 strategies for containing a lot. AI. And one of the easiest, it seems, is having more researchers working on safety. Do you have more researchers working on copilot safety? I mean, one thing that worries me is people using your AI to help them do destructive things or further their destructive views. Is that something you're thinking about at Microsoft? Yeah, we have a big safety team. We are definitely very focused on that. We're very focused, particularly on the, the tone of the AI. Like, how do we make sure that it isn't too sycophantic?

How do we make sure that it doesn't isn't over flattering? How do we make sure that it doesn't mirror you and sort of lead to this sort of negative cycle of reinforcing unhealthy views? And that's a real art and craft in trying to sort of engineer that healthy balance where, you know, your sort of AI companion can push back on your views in constructive ways without making you feel judged or making you feel angry, make you feel heard for your anger. It may be the case that you are angry about immigration, that you feel that you haven't had the opportunities and access to jobs in your community

that you feel have been available to new people coming into your world. And so, you know, it's about being respectful and acknowledging that people do genuinely feel aggrieved by that and not shutting them down because they don't adhere to some meta view. So, you know, and I think that's a very challenging line to draw. It requires, you know, real care and attention. So what role do you see yourself playing in terms of pushing the tech industry towards the public good? I mean, is that a role that you sort of are taking on? What do your fellow technologists think when they hear you talking about some of the more pessimistic visions you have for how AI could be deployed? Well, I think I'm both a pessimist and an optimist,

and that's not a bias. It's just an observation of the landscape before us. So most of all, I'm inspired by the practice of science. We have to say what we see and do our best to refute a hypothesis that we have with respect to evidence. And so I see evidence for both trends, and that's why I wrote about them. Look, I think, you know, in terms of how we're shaping the industry, I'm a big fan of the, you know, work that many of these, you know, sort of NGO organizations and social activists have been doing in order to raise questions and to challenge and push back.

I think that's healthy. We need more of that. And I'm very open-minded to it. I've been very sort of encouraging of additional regulation for a long time. I think this is a moment when going slowly and adding friction to the system will be long-term beneficial. And I think it's rational to just be a little cautious and increase the burden of proof. You know, and just make it a requirement that, for example, an AI shouldn't just be a straight, forward imitation of a human. We want to create an aid, a conciliaria tool that is an amplifier

and a supporter. And so, you know, there's kind of a lot of things to think through in terms of how this manifests in the world. Do you think there should be mandatory testing requirements that before any technology is released to the public, there's a, they should have to go through a certain series of tests. And if they don't pass, they don't make it to market. I think we're approaching that time. Yep. I think sometime before 2030, we will need something like that. I don't know if now is quite the right time.

If you look back on the impact that these sort of chatbots have had in the last two or three years, it's been unbelievably positive, right? It's sort of overwhelmingly positive, right? So had we had those checks ahead of time three years ago, I think it could have slowed things down quite a lot. But that doesn't mean that it's not right to keep asking that same question every year. And reconsider is now the right time for pre-deployment testing. I think that's the right question.

There is a sense that big tech needs to regain trust from consumers. There's a lot of people who just thrown up their hands and said, well, the convenience has outweighed all the digital privacy problems that we have. I give up. This is just the world we live in. But there are other people who are saying, I don't feel great about giving a tech company all my data, which is what you need to run some of these new AI tools, especially if you're telling me that terrible things could happen to it. You're clearly very cautious when it comes to the incredible powers of technology. but how are you balancing that with the demands on you to innovate and sell these products? Well, I personally think that there is going to be a huge amount of value to the user to have your co-pilot companion be able to read over your email, look at your calendar, schedule things, buy things for you, book and plan.

And, you know, I think the truth is we have to wait and see if consumers agree with that. They may not. And that's a very fair and reasonable thing to do. But I believe that the utility will drive the way. And in conjunction with that, we have to make sure that we have the strongest privacy and security infrastructure to protect that information, just as we already do today. Right. I mean, many, many billions of people store their email with Microsoft and rely on Microsoft's security to protect their consumer work and their enterprise. work. And so that's a massive, massive priority for the company. But, you know, it isn't just about

the utility. Obviously, it has to be really useful. Obviously, it has to have good privacy and security controls. But I think it's also about the way that we approach it. Like, how do we take feedback? And do we admit when we make mistakes? You know, how open-minded are we to different ways of doing things? You know, what's our business model? So I think my commitment is to sort of try to be as open mind as I can on all those questions and just listen and just innovate carefully, observe the consequences and iterate as we go. That's the best I can see to do at the moment. In the past, unlocking economic growth often came with huge downsides. The economy expanded as people discovered new continents and opened up new frontiers. But they colonised populations at the same time.

We built factories, but they were grim and dangerous places to work. We struck oil, but we polluted the planet. Now, because we are still designing and building AI, we have the potential and opportunity to do it better, radically better. And today, we're not discovering a new continent and plundering its resources. We're building one from scratch. Sometimes people say that data or chips

of the 21st century's new oil. But that's totally the wrong image. AI is to the mind what nuclear fusion is to energy. Limitless, abundant, world-changing. And AI really is different. And that means we have to think about it creatively and honestly. We have to push our analogies and our metaphors to the very limits to be able to grapple with what's coming.

because this is not just another invention. AI is itself an infinite inventor. And yes, this is exciting and promising and concerning and intriguing all at once. To be quite honest, it's pretty surreal. But step back, see it on the long view of glacial time, and these really are the very most appropriate metaphors that we have today.

Since the beginning of life on Earth, we've been evolving, changing, and then creating everything around us in our human world today. An AI isn't something outside of this story. In fact, it's the very opposite. It's the whole of everything that we have created, distilled down into something that we can all interact with and benefit from. It's a reflection of humanity across time.

And in this sense, it isn't. a new species at all. This is where the metaphors end. AI isn't separate. AI isn't even, in some senses, new. AI is us. It's all of us. And this is perhaps the most promising and vital thing of all.

As we build out AI, we can and must reflect all that is good, all that we love, all that is special about humanity, our empathy, our kindness, our curiosity, and our creativity. This, I would argue, is the greatest challenge of the 21st century, but also the most wonderful, inspiring, and hopeful opportunity for all of us. Thank you. That was Mustafa Suleiman.

He's the CEO of Microsoft AI and the author of the book The Coming Wave, Technology, Power, and the 21st Century's Greatest Dilemma. You can see his full talk at TED.com. Thank you so much for listening to our episode on The Future of AI. It was produced by Katie Montalione. It was edited by Sanaz-Mashkampore and me. Our production staff at NPR also includes Rachel Faulkner White, Harsha Nihada, James De LaHouci, Fiona, Fiona, our executive producer is Irene Noguchi.

Our audio engineers were Tiffany Vera Castro, Becky Brown, and Gilly Moon. Our theme music was written by Romteen Arablewey. Our partners at TED are Chris Anderson. Roxanne Highlash, Alejandra Salazar, and Daniela Ballorezzo. I'm Anoush Zamorodi, and you have been listening to The TED Radio Hour from NPR.