I’ve been lucky enough to have been working on AI for almost 15 years now. 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 intelligence, as your 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 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. It wasn’t long, though, before AI started beating humans at a whole range of tasks that people previously thought were way out of reach. Understanding images, translating languages, transcribing speech, playing Go and chess, and even diagnosing diseases. People started waking up to the fact that AI was going to have an enormous impact, and they were rightly asking technologists like me some pretty tough questions. Is it true that AI is going to solve the climate crisis? Will it make personalized education available to everyone? Does it mean we’ll all get a universal basic income and won’t have to work anymore? Should I be afraid? What does it mean for weapons and war? And, of course, will China win? Are we in a race? Are we headed for a mass misinformation apocalypse? All are good questions. But it was actually a simpler and much more fundamental question that left me puzzled. One that actually gets to the very heart of my work every day. One morning over breakfast, my six-year-old nephew Caspian was playing with Pi, the AI I created at my last company, Inflection. With a mouthful of scrambled eggs, he looked at me plain in the face and said, But Mustafa, what is an AI anyway? He’s such a sincere, curious, and optimistic little guy. He’d been talking to Bi about how cool it would be if one day in the future he could visit dinosaurs at the zoo, how he could make infinite amounts of chocolate at home, and why Pi couldn’t yet play iSpy. Well, as I said, it’s a clever piece of software that’s read most of the text on the open internet, and it can talk to you about anything you want. Right! So, like a person, then, I was stumped and genuinely left scratching my head. All my boring stock answers came rushing through my mind. No, but AI is just another general-purpose technology, like printing or steam. It’ll be a tool that will augment us and make us smarter and more productive. And when it gets better over time, it’ll be like an all-knowing oracle that will help us solve grand scientific challenges. You know, all of these responses started to feel, I guess, a little bit defensive and actually better suited to a policy seminar than breakfast with a no-nonsense six-year-old. Why am I hesitating? I thought to myself. You know, let’s be honest. My nephew was asking me a simple question that those of us in AI just don’t confront often enough.
What is it that we are actually creating? What does it mean to make something totally new and fundamentally different from any invention that we have known before? It is clear that we are at an inflection point in the history of humanity. On our current trajectory, we’re headed towards the emergence of something that we are all struggling to describe. And yet, we cannot control what we don’t understand. And so the metaphors, the mental models, the names—these all matter if we are to get the most out of AI while limiting its potential downsides. We should, I think, be able to easily describe what it is we are building, and that includes the six-year-olds. So it’s in that spirit that I offer up today the following metaphor for helping us to try to grapple with what this moment really is: I think AI should best be understood as something like a new digital species. Now, don’t take this too literally, but I predict that we’ll come to see them as digital companions, new partners in the journeys of all our lives. Whether you think we’re on a 10-, 20-, or 30-year path here, this is, in my view, the most accurate and fundamentally honest way of describing what’s actually coming. And above all, it enables everybody to prepare for and shape what comes next. Now, I totally get that this is a strong claim, and I’m going to explain to everyone, as best I can, why I’m making it. But first, let me just try to set the context. From the very first microscopic organisms, life on earth stretches back billions of years. Over that time, life evolved and diversified. Then, a few million years ago, something began to shift. After countless cycles of growth and adaptation, one of life’s branches began using tools, and that branch grew into us. We went on to produce a mesmerizing variety of tools. At first, slowly, and then with astonishing speed, we went from stone axes and fire to language, writing, and eventually industrial technologies. One invention unleashed a thousand more, and in time, we became homo technologicus. Around 80 years ago, another new branch of technology began. 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, and 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 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, images, music, and videos that stretch the imagination. People said it would never be empathetic. And yet today, millions of people enjoy meaningful conversations with AIs, 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 more computation than the deep-mind AI that beat the old-school Atari games just over 10 years ago. That’s nine orders of magnitude more computation. Ten X 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 eight billion words. And, of course, that’s a lot of words. But today, the most advanced AIs consume more than eight 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.
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 a near-perfect IQ. They’ll also have exceptional EQ. They’ll be kind, supportive, and 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, a doctor, a business strategist, and a coach are all in your pocket 24 hours a day. But things really start to change when they develop what I call AQ, their action quotient. This is their ability to actually get stuff done in the digital and physical worlds. And before long, it won’t just be people who have AIs. Strange as it may sound, every organization, from small businesses to non-profits to national governments, will have its own. Every town, building, and object will be represented by a unique, interactive persona. And these won’t just be mechanistic assistants; they’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 organizes a community get-together for an elderly neighbor, and a sympathetic expert helps 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, and 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 and take in every pattern of sensor data, sights, sounds, streams, and streams of information, faster-passing what any one of us could consume in a thousand lifetimes. To ensure that this new wave always serves and amplifies humanity, we need to find the right metaphors for what this might become. For years, we in the AI community, and I specifically, have had a tendency to refer to this as just tools. And we’ve seen that in the past. But that doesn’t really capture what’s actually happening here. 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 a new kind of digital species. Now, it’s just an analogy.
It’s not a literal description, and it’s not perfect. For starters, 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 memories. 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 are 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 it 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 and what AIs we bring into the world. Or not. These are the questions for all of us here today and all of us alive at this moment. For me, the benefits of this technology are stunningly obvious, and they inspire my life’s work every single day. But quite frankly, they’ll speak for themselves. Over the years, I’ve never shied away from highlighting risks and talking about downsides. Thinking in this way helps us focus on the huge challenges that lie ahead for all of us. But let’s be clear. There is no path to progress if we leave technology behind. The prize for all of civilization is immense. We need solutions in health care and education to our climate crisis, and if AI delivers just a fraction of its potential, the next decade is going to be the most productive in human history. Here’s another way to think about it: 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 also colonized 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 are 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, and world-changing. And AI really is different. 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, promising, concerning, and intriguing all at once. To be quite honest, it’s pretty surreal. But step back, see it in 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. Here’s what I’ll tell Caspian next time he asks. 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 that even a six-year-old can get a sense of. As we build out AI, we can and must reflect on 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. Thank you. Thank you, Mr. Stafford. It’s an amazing vision and a super-powerful metaphor. You’re in an amazing position right now. You are connected at the hip to the amazing work happening at OpenAI. You’re going to have resources made available. There are reports of these giant new data centers, 100 billion dollars invested, and so forth. And a new species can emerge from it. I mean, you’ve done it in your book, as well as painting an incredible, optimistic vision. You were super eloquent on the dangers of AI, and I’m just curious, from the view that you have now, what is it that most keeps you up at night? I think the great risk is that we get stuck in what I call the pessimism aversion trap. You know, we have to have the courage to confront the potential of dark scenarios in order to get the most out of all the benefits that we see. So the good news is that if you look at the last two or three years, there have been very, very few downsides, right? It’s very hard to say explicitly what harm an LLM has caused. But that doesn’t mean that’s what the trajectory’s going to be over the next 10 years. So I think if you pay attention to a few specific capabilities, take, for example, autonomy. Autonomy is obviously a threshold over which we increase risk in our society, and it’s something that we should step towards very, very closely. The other would be something like recursive self-improvement. If you allow the model to independently self-improve, update its own code, and explore an environment without oversight and, you know, without a human in control to change how it operates, that would obviously be more dangerous. But I think that we’re still some way away from that. I think it’s still a good five to ten years before we have to really confront that, but it’s time to start talking about it now. A digital species, unlike any biological species, can replicate not in nine months but in nine nanoseconds and produce an indefinite number of copies of itself, all of which have more power than we have in many ways. I mean, the possibility of unintended consequences seems pretty immense.
And isn’t it true that if a problem happens, it could happen in an hour? No. That is really not true. I think there’s no evidence to suggest that. And I think that’s often referred to as the intelligence explosion. And I think it is a theoretical, hypothetical maybe that we’re all kind of curious to explore, but there’s no evidence that we’re anywhere near anything like that. And I think it’s very important that we choose our words super carefully, because, as you’re right, that’s one of the weaknesses of the species framing. We will design the capability for self-replication into it if people choose to do that, and I would actually argue that we should not. That would be one of the dangerous capabilities that we should step back from. So there’s no chance that this will emerge accidentally. I really think that’s a very low probability. It will happen if engineers deliberately design those capabilities in and if they don’t make enough efforts to deliberately design them out. And so this is the point of being explicit and transparent about trying to introduce safety by design very early on.
