[0:00:00] What's up, everybody? My name is Dimitri Kofinas, and you're listening to Hidden Forces, a podcast that inspires investors, entrepreneurs, and everyday citizens to challenge consensus narratives and to learn how to think critically about the systems of power shaping our world. My guest on this episode of Hidden Forces is Mustafa Suleiman, a renowned tech expert and venture partner known for his development of leading AI models and as the co-founder.
[0:00:29] Google's Deep Mind. Our conversation today explores the potential consequences that stem from the rapid advancement and proliferation of artificial intelligence and synthetic biology, why this revolution is going to be fundamentally different than past periods of rapid technological change, and what can be done to contain these technologies in order to prevent some of the more catastrophic outcomes from occurring while still reaping the benefits of this next wave of innovation.
[0:00:57] We are approaching a critical threshold in the history of our species. Everything is about to change. At least, that seems to be the consensus that is emerging from my conversations with people like Mustafa, who work most closely with these technologies. And we spend the first hour of our conversation today wrapping our arms around not only how these systems work, how they learn, and how they perceive the world, but also why they are so difficult to contain and why we need to contain them.
[0:01:26] If we want to have any realistic hope of avoiding those catastrophic outcomes, things like synthetically engineered pandemics, engineered public confusion that undermines our sense of reality and moves either toward totalitarianism or toward the kind of anarchy that could generate civilization-ending terror attacks by lone actors capable of scaling destruction to levels that are difficult to comprehend and yet perfectly plausible.
[0:01:53] In the second hour, we discuss various versions of what success and failure might look like and what we can do as a society to help us move closer toward the future that we want and away from the dystopian nightmares that we've all seen depicted in some of the most famous movies of the last half century. You can access that part of the conversation at hiddenforces.io slash subscribe by joining our premium feed, which you can listen to on your mobile device using your favorite podcast app.
[0:02:22] Just like you're listening to this episode right now. If you want to join in on the conversation and become a member of the Hidden Forces Genius Community, which includes Q&A calls with guests, access to special research and analysis, in-person events and dinners, you can also do that on our subscriber page. And if you still have questions, feel free to send an email to info at hiddenforces.io and I or someone from our team will get right back to you.
[0:02:51] And with that, please enjoy this uniquely informative and mind-blowing conversation with my guest, Mustafa Suleiman. Mustafa Suleiman, welcome to Hidden Forces. Thanks for having me. It's great to be here. It's my pleasure having you on, Mustafa. So I was telling you that I finished reading your book about a month ago, and I went back to revisit the book and the questions that I had reached out, and I'm very excited for today's conversation.
[0:03:23] Into it. I'd love to know a little bit more about you. Just kind of a basic background. I mean, your bio speaks for itself, but people who don't know who you are, how would you describe yourself and what you do? You know, I mean, fundamentally, I'm an entrepreneur. So everything that I have done has been through that lens of building and starting things. In fact, it's pretty much the only thing I've done. I dropped out of Oxford to start a charity, which I helped run for three years, still going 20 years later. It's a mental health support service.
[0:03:53] I'm an atheist. I was an atheist as a late teenager, but I still saw value in that and really believed in the importance of providing faith and culturally sensitive support. I then co-founded a conflict resolution firm. I worked as a facilitator and negotiator all around the world, the age of 22, 23, 24.
[0:04:19] Gave me incredible insight and access in politics, in big business, in social organizing and activists. And it was just a hugely inspiring and eye-opening experience, both in terms of the limitations of what can be done at scale in the world and also the importance of technology and the lack thereof back in early 2000s. And then for the last 15 years of my career, I've been a startup entrepreneur, founding, scaling, creating.
[0:04:48] Two businesses now, DeepMind and then now Inflection over the last two years or so, specifically focused on AI. I've been trying to build general purpose learning systems that can essentially take what has made us successful, smart, creative, effective as humans, our intelligence, and reproduce that cheaply and ultimately more hopefully make it widely available to everybody. I think it's going to be the most transformative event of our lifetime.
[0:05:17] That's what I've been doing for the last 15 years. When I was growing up, there was basically, you were religious, you were an atheist, or you might be agnostic. That was kind of more of a fringy term that people began to use. What does it mean in your mind to be an atheist? Yeah, I think I'm quite a committed atheist and an intellectually committed atheist. I grew up very religious. My family were strict Muslim. My mom's English, and she was a convert.
[0:05:45] My father was from Syria. And just frankly, a very fundamentalist interpretation, very literalist interpretation. A materialist interpretation also of reality? Yep. Very traditionalist. And I think as I got to the age of 14, 15, 16, I started asking the fundamental questions about why are we the chosen people? Why is it the case that whilst on the one hand, Muslims are very proud of their...
[0:06:13] Post-racial equality, and often cite this as an example of being forward-thinking. You're united under one religious belief, regardless of your race. And that is great. But on the flip side, there's very little time given to LGBT people at all, if any. And the differences between men and women are objectively unfair and unjust.
[0:06:41] There is no real acceptable explanation for that as far as I can tell in scripture or in Hadith, in the kind of writings of the prophet and his followers afterwards. So those were very obvious inequalities to me when I was a kid and essentially led me on a quest to try to find alternative philosophies that seemed more just.
[0:07:02] And, you know, I discovered the human rights framework and I found that very inspiring. It's a foundational set of values that are, I think, mostly aspirational, but aspirational in important ways. You know, they provide a North Star for how we should interact with one another, regardless of, you know, race.
[0:07:22] Gender and sexuality and including religious belief. And we should be very accepting and supporting and respectful of people's right to choose religious belief. It just happens to not be mine. And the reason I'm not agnostic is just because I think it's a kind of philosophical get out. I mean, you can be agnostic on that basis about absolutely anything in the world, any fantasy belief or idea. You could technically say, well, there's no
[0:07:52] Proof that it's not the case, and therefore, philosophically, I probably should be agnostic. But I think that's kind of so reductive that it doesn't actually shed light on what we do know and what we can see, which is there's absolutely no evidence supporting any of the claims that come from the religious doctrine.
[0:08:10] And until we see some evidence for that, we should just assume that it's a non-starter rather than just engage in the kind of philosophical curiosity of there being a possible higher force. I'm open to that in principle, but that doesn't make me an agnostic. That just makes me philosophically curious. So maybe we'll have a chance to get into that later because epistemology and ontology...
[0:08:31] And ethics too are an important part of what goes into thinking about AI and AI safety. But before we do that, the name of your book is The Coming Wave, Technology Power, and the 21st Century's Greatest Dilemma. How did you go about choosing the title of the book? What's the metaphor of the wave here? And what do you mean by the 21st Century's Greatest Dilemma? So there have been general purpose technologies since the beginning of time. I mean,
[0:08:58] Fire and then the hand axe, you know, right the way through to the last century where we've had steam and electricity, the microchip and so on. What's different about this new wave of technologies, there's two new waves, artificial intelligence and synthetic biology, is that they have a series of characteristics which on the face of it make them seemingly uncontainable. They have these inevitable forces.
[0:09:27] that are driving their creation and that characterize what they actually are. For example, both have, by their very nature, autonomous characteristics. Once established, they have the capacity to adapt and integrate and update themselves and almost take on a life of their own. And we can talk about that in detail because I don't want to get too fantasy about it. But there are some fundamental characteristics of...
[0:09:56] And as a result, those of those characteristics, which you can talk about, they will by default proliferate far and wide. And I think that's been the history of all general purpose technologies so far is that to the extent that they are valuable, which all general purpose technologies are, they get cheaper and easier to use. And so they spread far and wide because everyone wants access and we find more efficient ways of making them.
[0:10:22] And that represents a dilemma because on the one hand, this is going to be the engine of progress that will make the 21st century and in fact, the next couple of decades, the most productive in the history of our species without question. I mean, we are about to see zero marginal cost everything, in my opinion, energy, food, materials, access to education, some aspects of healthcare. And I think that's going to be utterly transformational.
[0:10:51] But at the same time, that very force that is reducing the cost of everything over the next 30 to 50 years is also going to empower everybody with the ability to take action at scale. You can think of it as the plummeting cost of power. It's the name of the chapter in my book. It's really about everything, including the bad things, getting radically cheaper. And so the dilemma is treading this balance between the natural tendency that
[0:11:20] Many nation states will have to try and centralize and stop this proliferation of power versus the inevitable forces that are driving the proliferation and the spread of this power. And this isn't in many ways a new dilemma. This is the nature of the grand bargain of the nation state over a couple thousand years, that we centralize power in the public interest. And then there are moments when...
[0:11:45] That centralization weakens and power disperses for good reason. And generally when power disperses, we see great leaps forward. Everybody gets access to a new commodity or a tool or to a new insight or to a new culture. And that tends to be transformational. And so we should be incredibly optimistic about that upside.
[0:12:08] But at the same time, these are unique technologies which we haven't seen before, the ability to engineer new synthetic life biologically and artificially. And so that will bring fundamentally new characteristics as well. Right. And I do want to get into the upside and the downside. In the book, you actually write about how this is both a book about the advancements and expected advancements in AI and also in synthetic biology. In terms of artificial intelligence, what do we mean when we say,
[0:12:38] Artificial intelligence refers not to an entity or a thing, but to a field that is focused on trying to learn from what has made us successful and capable as a species. Our ability to, number one, synthesize information so we can absorb vast amounts of perceptual data visually, through audio, through touch, linguistically, and we can
[0:13:08] Find patterns in that information that allow us to make predictions about how some environment will unfold. I know that if I pick up the glass of water on my table and if I drop it off the edge, I can predict what's going to happen. I know that if I start shouting at you right now, I can predict to some extent what you might do.
[0:13:31] And we make predictions all the time. That's what humans are. They're sequential predictors. We don't just make one-shot predictions. We make predictions that unfold over a series of abstract components. So if I say, I want to predict what my journey would look like if I need to get to Delhi, I can imagine going home, filling my suitcase, getting an airline ticket, getting in a cab, going on a plane. All of those are individual.
[0:14:01] Predictions that string together to generate a series of predictions. And then I can use that trajectory of those predictions to intervene in that course. So if I'm being creative, for example, and I'm inventing something, I can imagine what that course looks like and I can tweak the trajectory of those predictions.
[0:14:22] to change the nature of the software I write or the painting that I make, or what might happen if I add this new fin or wing to a car that I'm designing. And very abstractly, that is really the essence of intelligence, taking information, making predictions, intervening in the course of those predictions. And so the science of artificial intelligence is to try to distill that process into
[0:14:48] A software engineering pipeline that we can use to achieve human level performance across a wide range of tasks. So I feel like, at least in everyday human experience, there are sort of two ways in which we make predictions. One is we intuit things based on... I mean, both have to do with past experience, but one is...
[0:15:10] It's intuitive. It's subconscious. We don't really have a model that we access and say, I understand this system. And based on these parameters, I have this set of predictions that I weigh accordingly. We kind of just react. And a good example would be in a fight. Fighters, for example, elite fighters are in a flow state using intuition throughout the course of the fight. And there is, of course, strategy and planning, which is different. But I think that captures sort of what I'm getting at here. And the other one is,
[0:15:40] A kind of modeling that ultimately stems from an understanding. I have an understanding of the system. I have an understanding of the procession of the equinoxes, and therefore I can make predictions about the positioning of the planets and the stars and everything else. How does that relate to the way in which AI makes predictions about the world and interacts with humans in the systems that we sort of understand to be AI driven thus far? Yeah, that's a great question actually.
[0:16:10] So I think the way to understand it is that if you look back over the last decade, so far, deep learning and then transformers have been principally focused on classification. They have been trying to understand- Tell our listeners what a transformer is before you, I just want to don't take their knowledge for granted. Sure. So deep learning has been the field of AI science that has powered pretty much all the progress of the last decade.
[0:16:40] And they are large neural networks where you sort of feed in raw data and the neural network learns some abstract representation of the components of that data. So if you feed in an image of a dog, the neural network will learn that pixels that are correlated in space represent some sub element of that image. Like for example, it could represent a claw.
[0:17:09] And then at a smaller abstraction of pixels, it might represent a straight line or a curved line or an edge, a corner. And you can actually create these sub-representations of commonly occurring jigsaw pieces of the puzzle that can then be layered up in a hierarchical way, all the way up to the idea of sort of leg.
[0:17:35] And then body, and then the full dog, and then the scene that the dog's in, etc, etc. And that's been the classification revolution of the last decade. And that's how we've managed to get deep learning models to be able to achieve pretty much human level performance on recognizing faces, understanding scenes, images, understanding audio so that you can have transcription on your phone, and increasingly understanding some of the content of text as well.
[0:18:05] But the flip side of that is once a model understands the content of raw data well enough, you can then ask it to generate a new example of the class object that it understands. So it knows the difference between cats and dogs and zebras and horses so well that you can say, well, generate me a zebra that looks like a cat with some dragon wings in a Pokemon style.
[0:18:32] And that's essentially doing an interpolation. It's taking this multidimensional space of representations that it knows and finding the joint representation that best fits those five or six, or in many cases, many millions of dimensional points that represent that object. Statistical variations of data sets that it's seen before that are sufficiently similar, is it based on a structural understanding of the data or a statistical understanding of the data?
[0:19:01] And not necessarily a derivation of meaning from the data. In other words, can we separate those two? Do we understand whether it actually understands it the way that we understand something? Well, I mean, the idea of understanding is a very flaky, in my opinion, abstract, somewhat poorly defined notion. It's not quite as bad as consciousness, but it is on that trajectory. We don't know what it means to truly understand something.
[0:19:26] I don't know that you understand something. I just know that from what you say and what I can observe about your actions, I can verify that with respect to what is true, what is effective, what is valuable. I'm a behaviorist in that sense. Whether or not you understand, I can only evaluate by the behaviors that you exhibit.
[0:19:47] And I actually think that's the correct standard to evaluate the progress of these models to. We can come onto the idea of artificial capable intelligence, but really what we should be doing is measuring capabilities, what an AI can do, rather than whether it has deep understanding or whether it has consciousness, because these are very sort of slippery concepts that we haven't yet been able to fully grasp. Good. So let's talk about that. What is the difference between artificial general intelligence, AGI, and artificial
[0:20:17] Capable intelligence, ACI. Well, so artificial general intelligence is the super intelligence that people sometimes characterize. An AI that can perform every human task better than a human and will use that to bootstrap itself up to an even higher intelligence. It will keep getting better and better. Runaway AI or something like that. That's right.
[0:20:46] A recursively self-improving AI. And that's the kind of existential threat that people often talk about. Personally, I think that threat is sort of overstated. I think we're quite a long way from those kinds of risks. It's theoretically plausible. Like I said, kind of at the beginning with agnosticism, many things are theoretically plausible. And I think if we over-fixate on these sort of tiny probability outcomes...
[0:21:13] We missed the wood for the trees. So personally, I think a much more practical sort of area of focus is to say, let's focus on what the AI can do. We can measure that very practically and we can observe its outputs and we can use that to control it and contain it and make sure it remains accountable to humans and that it doesn't wander off and do things that are potentially dangerous. And so I've proposed a new version of the Turing test.
[0:21:40] The Turing test was first proposed in the 50s by Alan Turing, a famous British computer scientist. And he basically said, a machine is intelligent if when a person is talking to it, that person can't tell whether the conversation is with a human or with another machine. And it was really known as the imitation game. Just by saying the right thing at the right time, that should indicate that the machine is intelligent.
[0:22:08] In fact, now that we do have AIs, my own included at Inflection, our AI is called Pi, which stands for personal intelligence. It's an incredibly high emotionally intelligent conversational companion. And we don't really think it's intelligent. We don't know much more about its intelligence, but it's certainly an amazing conversationalist. Incredible. Very, very engaging. You can actually telephone it and speak to it fluently on the phone.
[0:22:36] As you would another human, just as I'm talking to you now. So it's remarkably high quality. And yet we haven't learned much more about whether it's intelligent or not. So I think that a modern version of the Turing test is one that instead evaluates what the AI can do compared to a human. So for example, can it make plans? You mentioned planning earlier. Planning is an extremely hard skill.
[0:23:04] You have to string together a sequence of predictions, all of which are accurate and relate to one another correctly over time. Can it learn to use tools? Can it use APIs, application programming interfaces? Can it access information from databases? Can it telephone other humans and ask for the key piece of information that the AI is missing? Can it phone other AIs?
[0:23:28] So there's all sorts of very practical capabilities that we can use to evaluate its performance and what it can do in the world. So if we were having this conversation, let's say in the early 1990s, and we were talking about the early internet and web protocols and the opportunities for the commercial applications and some of the stuff that could come out of it, that's a sort of a conversation that most people can understand with a new technology, right? It's like all the things that it can do. And how does a conversation about AI...
[0:23:57] Differ from that framing. And what in your view is at stake that compelled you to write this book? Because this book is not about, it's not even primarily about the opportunities. I mean, you talk about them and there are a lot of exciting opportunities. And I feel like a lot of times what ends up happening in our conversations about technology is we focus, social media is a great example, and I do this as well. We focus on all the really bad ways in which the technology is used. And I think those are warranted. And we forget all the really great ways in which they are used.
[0:24:26] When it comes to AI and synthetic biology and additional technologies that we could expect to come out of the future that we're moving towards, there are a lot of really great opportunities. But again, you focus a lot of your time on the book on the downside risks associated with that. So how should people frame this conversation in their heads to think about really what's at stake here and how it's different than previous technologies? So I think the simplest framing is that intelligence is what has made us
[0:24:54] Civilized, wealthy, healthy caused the population boom of the last 150 years. And intelligence isn't some mystical abstract idea like understanding or consciousness, but it's actually something that we can engineer. We can engineer models to make predictions which produce capabilities, which look like...
[0:25:21] Creativity, which look like empathy, which look like judgment, which enable planning, even reasoning, deductive reasoning, inductive reasoning, logic. Capabilities that we may have thought would be the preserve only of humans. In fact, people have been saying this constantly for as long as I've been in this field for 15 years. Two or three years ago, people were like, AIs will never be creative. Humans are always going to be creative, right?
[0:25:50] And then obviously we've objectively seen that they're incredibly creative as tools. A few years before that, people say, oh, well, AI is never going to be empathetic. Kindness, compassion, gentility, support. It's always going to be something that humans can do. And I think that's also largely proven not to be the case. I mean, AIs will be more consistent, more stable, more reliable.
[0:26:18] Less judgmental, more supportive, more fair, more kind. So the upsides are truly incredible. Your AI scientist or your AI researcher is going to be less biased, more accurate. It isn't going to be prone to hallucinations. It's basically going to be near perfect. Now, the downside of perfection is that some people will choose to use this to sow massive instability.
[0:26:47] Just as people have used iPhones and hammers and every tool that has been invented ends up getting used for darkness as well as the light. What's different this time about the technology is that power is proliferating at an unprecedented rate, which means smaller groups of bad actors.
[0:27:16] with fewer resources are going to have an easier time of reaching millions of people. Just as in the last wave of software, what happened? We had open access to the internet, which means that anyone can publish. Suddenly we've got this new format called podcasts. Anyone can write a blog. Anyone can post on social media. Mainstream media is undermined and slowed down, largely for good reason, I think. The information ecosystem is more open.
[0:27:45] Yes, of course, it's more polluted, but individuals with no formal training, no network access, no access to capital, access to power, can now broadcast. Anyone can have a one-to-many broadcast effect if they choose. That same trend is about to happen, not just with respect to access to broadcasting information, but with taking actions.
[0:28:09] You're now going to have an ACI, an artificial capable intelligence that can do things for you to other people at huge scale with almost zero marginal cost, certainly over a 20-year period. And I think that's the trajectory that we're on, is zero marginal cost access to power. So I have another question about that, but there's a part in the book where you talk about how the coming wave launches these concurrent centralizing and decentralizing.
[0:28:34] Yeah, that's a great point. And I mean, I have an entire chapter focused on the techno-authoritarianism.
[0:29:00] trend, which is also inevitable. And actually, you don't have to be that creative. You can just take a look at China. You can take a look at some of the Gulf nations that have adopted large-scale surveillance systems. And it's going to be pretty clear that... So we were sort of told this myth in the 90s and 2000s that the web was inherently an open system and would always remain an open system. And actually, that's not true.
[0:29:29] There's nothing inherent about that. That's a design choice. If we want openness, we have to actually fundamentally protect that openness deliberately because we use these metaphors like web and cloud and stuff like that, but actually it runs on copper wires, fiber optic wires, and that information, those bits sit in data centers, which are subject to national law.
[0:29:57] There is a central source of truth which can be surveilled, which provides a very dangerous dystopian opportunity for authoritarians to try to censor and restrict and monitor people. I think that's going to get easier and easier because these tools definitely enable us to make sense of vast amounts of data. As we saw from the Snowden revelations, every country
[0:30:26] Even the countries in the world that see most on the face of it committed to openness and freedom and liberty, rightly so, will naturally have the impulse to try to collect it all. Some of the names of those programs, you know, Prism and Tempora and these collect it all efforts, you know, are pretty scary.
[0:30:46] They're pretty scary. And now the only thing holding them back then was that they didn't have algorithms that were capable of making sense of that information and finding the quote unquote needle in the haystack. Whereas these days we are going to have that. And of course, they're going to be widely available in open source. I think that's the kind of counterintuitive thing. That's the contradictory thing, both centralizing forces and decentralizing forces simultaneously. When you think about the future, like the expected arc of your life,
[0:31:16] How does that make you feel in terms of what you envision it looking like, where it's going? What are the feelings that come up for you as a person? Well, as I think you can perhaps see from the book, I'm anxious. I'm deeply, deeply worried. And I opened the book with this idea of pessimism aversion and the trap of having a default instinct, which is to look away from negativity.
[0:31:46] I think it's a particularly American value, actually. I think there's a huge optimism bias here. And I love that. I moved to Palo Alto because it's the center of the technology universe because everybody is energized by vitamin D and blue skies all day long. And it's amazing. People are fearless. People love failure. People celebrate success. It's just an amazing energizing environment to be in and completely different to Europe or London, for that matter, where I was born and raised. So I love that. But at the same time, there is a fear.
[0:32:15] So I don't want to mix my metaphors here because I actually wrote down a question for you heading into this interview, asking if we had gotten to a Sputnik moment yet with this technology. But in your response, I actually...
[0:32:45] Think about the period, really the 1950s, and this was shortly after the use of the atom bomb over Japan, and people had a real fear about the end of the world. There was a real concern, and it was pretty widespread, I don't know the number, but you could tell from the cultural artifacts of the time that a lot of people thought that there was a really good chance that we could blow up the world.
[0:33:13] There's the great scene of Charlton Heston kneeling in the sand in Planet of the Apes saying, as you maniacs, you did it. What is the appropriate level of concern that people should have today? And how does it compare to what we had then? And then sort of like what was appropriate then versus what we felt? And then what is appropriate now in your view versus how we feel?
[0:33:38] I think the idea that we're going to be surrounded by super intelligences which don't care about us and end up accidentally killing us all, this is completely misguided. And I think it's a distraction. I think there are much, much more consequential near-term risks which amplify the fragility that we already are suffering in the world. Democracy is on the ropes.
[0:34:06] Our institutions, political institutions and political figures have in the Western world never been distrusted more. And what's coming is going to be a massive shock to that system as anybody can create high quality synthetic media, which is personalized to sell whatever narrative they believe to be true. And I don't think that...
[0:34:35] It's really just about misinformation. Misinformation is too much of a catch-all term, which kind of doesn't really grasp at what's actually going to happen. What's more likely is, we all believe we're right. Even the quote-unquote bad actors are doing it because they're motivated by something that they believe to be a higher purpose. So what this is going to do is amplify that chaos because it's going to make it easier for everybody to send
[0:35:03] You know, new kinds of information which support their worldview into the information ecosystem and cause a kind of information apocalypse. It'll be very hard to tell what's what. Now, on the plus side, we're an incredibly resilient and adaptive species. In fact, that is the essence of what makes us human, coming onto what really is unique about us. We are able to very quickly discern
[0:35:32] What caused what and change and adapt extremely fast. Just look at the adaptation to chat GPT, for example, or look at our adaptation to COVID. Yes, there were a horrific number of avoidable deaths, but on the other hand, there's a huge achievement, a scientific and technical achievement.
[0:35:53] And in some respects, also a governance achievement as society didn't fall apart. Supply chains did actually stand up. People weren't cut off from food and water. So there are reasons to be very optimistic about our ability to quickly adapt to these dangers and find ways to prevent their biggest harms. And I think that we're going to have to expect that a lot of content online will be produced.
[0:36:20] So this wasn't the initial area where I was going. I was actually going to ask you about manufactured pandemics that could arise from synthetic biology, because that was something that I think I either heard you in an episode, another episode, another conversation that you had.
[0:36:45] On a different podcast or it was in the book where you talked about it or both where this is something that you could see happening in the next few years. But let's put a pin in that for a moment because the information problem is at the very top of my list and it's been at the very top of my list for a long time in part because I spend all my time trying to figure out what's true and what's not true and I struggle with it. So you brought up the COVID-19 pandemic and the successes around it. I also look at the failures around it and I have to say in terms of my own
[0:37:16] Effort on this podcast to be diligent around how I approach the pandemic. I had, for example, the former FDA commissioner, Scott Gottlieb on the program. We talked about the mRNA vaccines and the exciting technology that underlies them. And I've also, I have to say, I've only given this so much attention because I feel overwhelmed by the debate around adverse vaccine effects. What in fact, for example, Pfizer did in their trial data.
[0:37:45] What we know about whether or not the vaccines that they use in their trial data are actually the vaccines they deployed. There's such an onslaught of information. And so even someone like me, who I think is fairly intelligent and fairly informed and very well-intentioned, struggle to really know what is true and what is not true around something this fundamental. And so when I look going forward, and I look now at the Ukraine war, because again, on this podcast, we cover a lot of things and geopolitics is one of them.
[0:38:15] I watched how the Biden administration spent months trying to alert the public and the Europeans and Zelensky to this impending invasion of Ukraine. And what I saw was what consistently happens in this country and has been happening now increasingly over the years, which is a river turning into a series of streams and people just beginning to go off into their own universes and those beginning to develop their own narratives and evolving accordingly. Now,
[0:38:45] We're in a similar situation with Israel, the Levant, and the possibility of a regional war there. And what I feel like is happening, if you think about human beings, just animals or apes or whatever, we are hierarchical creatures and we look for leadership and we look for direction in order to organize in space. So I feel like the problem isn't just that we're in this really perilous place today in terms of not being able to know.
[0:39:12] What's true and what's not true on some very fundamental questions such that even me, where my job is this, struggle. But I feel like we're also beginning to lose faith in a very basic level in the institutional power structures of society. So that when we talk about AI and we talk about power and its distribution,
[0:39:33] I feel like, and you kind of touch on this in the book, because I don't remember if you actually talk about the Westphalian order, but in the book, you kind of hint at something that seems a little bit like we could be on the verge of a revolution on the scale of what we saw 500 years ago with the development of the nation state and this sort of organizing principle that states should have sovereignty over physical spaces. And I feel like all of this is coming apart. And I don't know...
[0:40:02] How to wrap my arms around it. So go ahead and just take that however you want. No, I mean, it's a fantastic point. That's exactly what I'm getting at in the book. And I think that part of the diagnosis has to be that we have started with a false expectation that we are rational.
[0:40:22] That we have established civilization. The language that somehow became so popular of the end of history in the 90s with Frank Fukuyama, I think led us to some very dangerous places because it sort of assumed there's this implicit belief that liberal democracy was this stable, lasting thing and that having struck upon it, it didn't need constant protection and that we had just arrived at this happy place.
[0:40:52] And in fact, truth is costly. Truth is expensive. It's exhausting. It's time-consuming. It's a constant battle. There is no stable end state. So the tributaries that you describe of sub-roots of information flowing in different pockets, that's actually just a curse of progress. And it's actually an amazing thing.
[0:41:19] The fact that we now have digitized information means that everybody gets access and we're just adapting to the last wave of the transistor microprocessors and the web. And we're coping with this new reality where even the least educated person on the planet has orders of magnitude more information at their disposal than the most educated people 100 years ago.
[0:41:48] And so as a species, we have to just cope with the fact that we can see evil. We can see complexity in a richer and more destabilizing light than has ever been possible. And that's just very, very hard to accept. And especially when we have schooled ourselves for centuries in the narrative of basically man, patriarchy.
[0:42:15] And religion, which is very simplistic. There is right or wrong. It is very clear. We know everything. The whole world revolves around our planet. We're still moving through that enlightenment period where we're coming to know how little we know and how overwhelmingly complex things are. And that's the journey that we're going through with technology, is that it is amplifying that fundamental fragility in our worldview.
[0:42:44] So I don't know if I want to sort of go down the information landscape rabbit hole because it's not really what the book's about. So I just have maybe one more question around that. Since you work in this space, you probably know all the big players. And when I say this space, I mean just tech in general. It doesn't make any sense to me that social media platforms are not regulated with the same seriousness that we regulated the banking system after the Great Depression. In other words, I don't understand why we haven't set clear rules of the road around transparency.
[0:43:12] Around the same kind of work that's being done on AI safety, I don't understand why that isn't being done and implemented in social media, because this is literally how people are coming to know the world, and people really are going crazy. And that's not a value statement. It's not a political statement. It's not me saying that I know what is the canonical truth. It's saying, in fact, in many cases, I don't know, and that's a problem. And I'm not even talking about...
[0:43:39] Ontological truth. I'm talking about epistemic truth. I'm talking about consensus truth. And so why is this problem not being addressed with greater urgency by people in the industry and people in government? What explains that? Look, I think the bottom line is the platforms were never platforms. The word platform implied this false notion of neutrality.
[0:44:03] As though it was just like a yellow pages index. And I'm not responsible for it. You make the phone call and you see if it's a good plumber. At least now everybody's familiar with that problem. But I think that was a big trip up of the last 20 years on social media and big tech. I mean, those companies are constantly doing moderation.
[0:44:24] Constantly editing, removing, amplifying, de-amplifying. So I think that now everyone sort of grasped that. Now it's very obvious that this should be sort of regulated, if not as popular. But they can't be trusted to do that. They can't be trusted to do that as a private corporation because it's too important. The public space is too important. And people don't trust them. People don't trust them. No, I totally agree. I'm not saying they should. I'm not saying they should. I'm saying that they're already doing it. So we should regulate the way that it's happening, right?
[0:44:54] Yeah. I mean, I completely agree with that. My question is why? So I've noticed reading the book and listening to your interviews that you take this problem very seriously. And when I say the problem, I mean AI. You take it very seriously. What I don't understand is, is there not a commensurate seriousness with which people in the industry take the information problem? Well, I think there's confirmation bias. So I think that they've built trillion dollar businesses with a certain business model.
[0:45:22] And they're sort of stuck in selling that story and they can't pivot away from, or they very, very rarely acknowledge the dangers and weaknesses of that business model because there's a structural advantage in keeping that narrative up, I think. I can understand why Mark Zuckerberg might feel that way or Elon Musk because he owns Twitter. What I don't understand, and maybe this is naive, and actually I'm not even saying it's naive because I'm not suggesting that they should. I guess I don't understand the incentives.
[0:45:49] Of this new network of very intelligent, financially successful people in tech who exert a significant amount of power comparable to the power of the robber barons in the early 20th century. What is it about that network that hasn't allowed or hasn't created the manifestation of voices to come out?
[0:46:13] Speak on behalf of the humanistic principles that you seem to value in the course of writing your book. Because we're in a place of desperate... We desperately need leadership. And guys like you and your peers are where a lot of the power is and where a lot of the decisions are being made that are going to affect the future of humanity.
[0:46:33] And so again, I don't want to go down this road on the information thing and we can edit here and kind of move over to the AI stuff, but it's the most immediate tangible thing. It's like the lowest hanging fruit in the world, the biggest bang for the buck in my view, from what I see in terms of right now, stuff that's causing havoc in our discourse, where we have one open-ended war in Eastern Europe. We have another potential war in the Middle East that's going to drag the United States in, that's going to completely have to chain the strategic pivot towards China. I mean, all this stuff is in disarray and we can't seem to agree on the most basic things.
[0:47:03] And I just, I don't quite understand it. Anyway, if you have any thoughts about it. Look, I'll give you two things that I can think of. One is that it's just incredibly hard to do moderation, right? And I think that that does need to be given some airtime, right? It is technically very, very difficult to manage billions of messages a day on video, on audio, et cetera, et cetera. However-
[0:47:31] The flip side argument to that is where there is an obvious and pressing need, the technical issue can easily be solved. And so when you look back in hindsight, it should have been made a priority beforehand. So for example, the Christchurch shooting massacre that was live streamed on Facebook so horrifically in New Zealand some years ago, I forget, five or six years ago now, that triggered a complete sea change in the amount of attention that the big companies were focused on.
[0:48:00] Trying to detect a live streaming of a mass shooting or some kind of equivalent harm. And they fixed it. They got it down to now like, I don't know, what is it? Sub few seconds, two or three seconds that it gets dealt with. And that's just like one example of many where the technical issues can be overcome if people are really prioritizing and focusing on it. I think the second big explanation is that...
[0:48:22] For as long as forever, big tech was the darling of the world, right? I mean, it could do no harm throughout the 2000s and early 2000s, mid 2000s. It really was the engine of progress and people were sort of in love with it. And it was like a new silver bullet. And again, it goes with this kind of end of history thing. I mean, it looked like we're kind of just solving on all fronts. I mean, you remember that sort of in the early 2000s and in the late 90s, Clinton and Blair would go around saying that this is the end of boom and bust.
[0:48:52] We've solved it. We've got this central banking system. We've solved the business cycle. Yeah, we've sold the business cycle and we've got interest rates under control. We've got inflation at 2%. It's all just technocratic. We're going to take care of it. Don't worry. So there's this complacency and arrogance that comes, in my opinion, from a post-religious era.
[0:49:11] Just to state the obvious, it's also men kind of thinking like, we're the bosses, we've got this, we're the patriarchs and so on. And just lacking humility that actually we can be evil, we can do wrong, we can make mistakes. And instead of starting from a position of assuming that we're actually fundamentally fallible, like Hannah Arendt style, which is what I believe, that there's evil inside all of us and it will come out if you're put in the right context and starting from that level of needing self-discipline around that.
[0:49:41] Then I think Silicon Valley and the source of the big tech companies started from a position of like, well, we're the good guys. We're the good guys. We wouldn't do anything wrong. And so then your blinkers are on and your attention is not focused on the potential harm that you're causing. And even when you encounter that harm, like for example, the constant talk of sort of bullying and anorexia stuff on Instagram and so on, that was...
[0:50:07] Very obvious to many, many parents and many people and many activists and so on for many years before Instagram took any action on that. And you think about the amount of avoidable harm. So I think it's just like level setting people on the potential harms and taking more proactive responsibility for the consequences of our actions. So Mustafa, I'm going to move us to the second hour. And that's where I want to get into solutions, the solutions that you propose in the book. And I also want to understand...
[0:50:34] What success looks like to you, what the threshold for success is, where let's say you look back 30 years from now or whatever the period is, 30, 40 years, and you say like, you know what? This isn't ideally what I wanted, but I'm satisfied. I'm okay with this outcome. And then what does failure look like? So that we understand kind of what the potential outcomes are here, and then also what we can do to push the vector of change in the direction that we want.
[0:51:03] For anyone who is new to the program, Hidden Forces is listener supported. We don't accept advertisers or commercial sponsors. The entire show is funded from top to bottom by listeners like you. If you want access to the second hour of today's conversation with Mustafa, head over to hiddenforces.io slash subscribe and sign up to one of our three content tiers. All subscribers gain access to our premium feed, which you can use to listen to the rest of today's conversation on your mobile device using your favorite podcast app.
[0:51:31] Just like you're listening to this episode right now. Mustafa, stick around. We're going to move the rest of our conversation onto the premium feed. If you want to listen in on the rest of today's conversation, head over to hiddenforces.io slash subscribe and join our premium feed. If you want to join in on the conversation and become a member of the Hidden Forces Genius community, you can also do that through our subscriber page. Today's episode was produced by me and edited by Stylianos Nicolaou.
[0:52:02] For more episodes, you can check out our website at hiddenforces.io. You can follow me on Twitter at Kofinas, and you can email me at info at hiddenforces.io. As always, thanks for listening. We'll see you next time.