All Segments
Balaji Srinivasan: How AI Will Change Politics, War, and Money
45 segments available
a16z General Partners Erik Torenberg and Martin Casado sit down with technologist and founder Balaji Srinivasan to explore how the metaphors we use to describe AI—whether as god, swarm, tool, or oracle—reveal as much about us as they do about the technology itself. Balaji, best known for his work in crypto and network states, also brings a deep background in machine learning. Together, the trio unpacks the evolution of AI discourse, from monotheistic visions of a singular AGI to polytheistic interpretations shaped by culture and context. They debate the practical and philosophical: the current limits of AI, why prompts function like high-dimensional programs, and what it really takes to “close the loop” in AI reasoning. This is a systems-level conversation on belief, control, infrastructure, and the architectures that might govern future societies. Timecodes: 0:00 Introduction 0:37 Personal Journeys in AI and Crypto 3:54 Monotheistic vs. Polytheistic AGI: Competing Paradigms 7:53 The Limits of AI: Chaos, Turbulence, and Predictability 9:36 Platonic Ideals and Real-World Systems 14:10 Surprises in AI Progress: Language, Locomotion, and Double Descent 25:45 Prompting, Verification, and the Age of the Phrase 29:18 AI, Crypto, and the Grounding Problem 34:26 Visual vs. Verbal: Where AI Excels and Struggles 37:19 The Challenge of Markets, Politics, and Adversarial Systems 40:11 Amplified Intelligence: AI as a Force Multiplier 43:37 The Polytheistic Counterargument: Convergence and Specialization 48:17 AI’s Impact on Jobs: Specialists, Generalists, and the Future of Work 57:36 Security, Drones, and Digital Borders 1:03:41 AI, Power, and the Balance of Control 1:06:27 The Coming Anti-AI Backlash 1:09:10 Global Implications: Labor, Politics, and the Future Resources Find Balaji on X: https://x.com/balajis Find Martin on X: https://x.com/martin_casado Stay Updated: Let us know what you think: https://ratethispodcast.com/a16z Find a16z on Twitter: https://twitter.com/a16z Find a16z on LinkedIn: https://www.linkedin.com/company/a16z Subscribe on your favorite podcast app: https://a16z.simplecast.com/ Follow our host: https://x.com/eriktorenberg Please note that the content here is for informational purposes only; should NOT be taken as legal, business, tax, or investment advice or be used to evaluate any investment or security; and is not directed at any investors or potential investors in any a16z fund. a16z and its affiliates may maintain investments in the companies discussed. For more details please see a16z.com/disclosures.
Segments Timeline
Polytheistic AGI: A Cultural Perspective
Balaji Srinivasan introduces the concept of polytheistic AGI, suggesting that every culture will develop its own version of artificial general intelligence (AGI), social networks, and cryptocurrencies. He argues that these technologies will serve as the core of modern internet societies, functioning as oracles that guide cultural values and social structures.
"So polytheistic AGI I think is one very useful macro frame means every culture has their own AGI and eventually every culture has their own social network and cryptocurrency and AI. You know the AI is..."
Personal Journeys in AI and Crypto
Balaji shares his personal journey in the fields of AI and cryptocurrency, detailing his academic background in machine learning and computational statistics. He reflects on the evolution of AI technologies and his transition from genomics to crypto during the deep learning revolution, highlighting the challenges of staying at the cutting edge of multiple fields.
"Martine and I were talking off cuh offline about how amazing your thread was on AI and uh you know normally or often you're a crypto guy, you're a network state guy, but you know you're you're you're ..."
Surprises in AI Progress
Balaji discusses his initial skepticism about the capabilities of AI, particularly in relation to models like GPT-2. He expresses surprise at the advancements in AI coherence and functionality, particularly with models like ChatGPT and DALL-E, which exceeded his expectations for AI's potential.
"just as a deep learning revolution was getting underway with ImageNet and um you know all all that series of papers in the mid2010s early 2010s and um so I've I've so just my thought process I have I ..."
Limits of AI: Chaos and Turbulence
In this segment, Balaji outlines the inherent limitations of AI, emphasizing that chaotic and turbulent systems cannot be predicted indefinitely. He explains how finite precision arithmetic and cryptographic equations impose bounds on AI's predictive capabilities, challenging the notion of AI as an all-knowing entity.
"coherent chat GPT was. I think everybody was, but it was like a huge jump up from what it was before in terms of sort of being Markoff chain. And so I've been kind of observing over the last, you know..."
The Monotheistic View of AGI
Balaji critiques the monotheistic perspective of AGI, which envisions a singular, all-powerful intelligence. He contrasts this with his polytheistic view, suggesting that multiple superhuman intelligences, each shaped by different cultural values, will emerge instead.
"come to more recently, but let me kind of enumerate them in no particular order. Okay, I made Martine jump in any time. So the first is that something that motivated uh you know I think both Ellie Eis..."
Cultural Variations in AI Development
Balaji elaborates on the idea that different cultures will produce distinct AIs, such as American and Chinese AIs, each reflecting their respective societal values. He discusses the potential for decentralized, open-source AI models to emerge, which could diversify the landscape of AI technologies.
"thinking so much about crypto and other kinds of things I was like well there's a different sort of implicit school of thought which is polytheistic agi right do we have rather than the vengeful god d..."
AI Apocalypse: A Misconception
Balaji addresses the fear of an AI apocalypse, arguing that current AI technologies, such as image generators and chatbots, are unlikely to lead to catastrophic outcomes. He emphasizes the importance of understanding AI's limitations and the context in which these technologies operate.
"That's a big thing like all the deepseek models etc. So polytheistic AGI I think is one very useful macro frame which takes away some of the sort of um AI apocalypse tones I think uh because I don't t..."
The Role of Cultural Context in AI
In this segment, Balaji discusses how cultural context shapes the development and application of AI technologies. He posits that each culture's unique values and norms will influence the design and functionality of their respective AIs, leading to a diverse array of systems.
"of a modern internet for society, right? Um and they'll be customized for each different kind of group and certain things will be disallowed and allowed like image generation might not be allowed in s..."
Understanding AI as System Software
Martin Casado contributes to the discussion by framing AI as system software rather than divine entities. He critiques the anthropomorphic view of AI and emphasizes the importance of recognizing the limitations of AI systems, particularly in relation to chaotic and turbulent phenomena.
"cogitate for millions of years and figure things out and it could could outmaneuver you all the time. And we know that's not true because turbulence, chaos, cryptographic equations are not like that, ..."
The Platonic Ideal of AI
The conversation shifts to the concept of the Platonic ideal of AI, with Balaji and Martin discussing how thought experiments have shaped public perception of AI. They caution against conflating these ideals with the realities of existing AI systems, highlighting the need for a pragmatic understanding of AI's capabilities.
"Can I can I can I just can I just add just a little bit of color here? I I think this is great. I think we need to call out why you know so the the way that you describe AI as like gods and monotheist..."
Defining AI Limitations
Balaji and Martin explore the limitations of AI, particularly in terms of decentralized models and the continuous evolution of AI technologies. They discuss how these limitations challenge the notion of a singular AGI and emphasize the importance of recognizing the real-world constraints faced by AI systems.
"problem with doing that, the problem with taking some Platonic ideal, whether it's Bostrums or it's whatever Abraham the the the Abrahamic view of God or or or or any kind of religious view is that it..."
The Role of Platonic Ideals in AI
Balaji Srinivasan discusses the significance of Platonic ideals in the context of AI, referencing thought experiments like the Turing Test and the Chinese Room. He emphasizes the importance of distinguishing between theoretical concepts and real-world AI systems, highlighting how these ideals can shape our understanding of AI's capabilities and limitations.
"that was a platonic ideal no no reference to neural networks no reference to implementation details and yet it served as something that went from a thought experiment to an applied thing with you know..."
Decentralized AI vs. AGI
In this segment, Balaji contrasts decentralized AI with the concept of Artificial General Intelligence (AGI). He argues that the rapid development of new AI models challenges the notion of a singular AGI, suggesting that the evolution of AI is more continuous than previously thought, and discusses the implications of this shift for our understanding of AI's potential.
"So so now your point which is a very good one is these are real systems and they actually have real limitations. I think one of the more interesting things for me over the last few, you know, years, m..."
Surprising AI Capabilities: Language vs. Locomotion
Balaji shares insights on the unexpected capabilities of AI, particularly in language processing compared to physical tasks like locomotion. He explains how AI can outperform humans in creative tasks such as writing, while still struggling with complex physical navigation, revealing the counterintuitive nature of AI's strengths and weaknesses.
"like to you know can can an AI write a sonnet? It can right it can do it better than most humans. Can it can it write a screenplay? It can do that again better than most humans. There's a lot of thing..."
The Evolution of Language Models
This segment explores the advancements in language models, particularly with GPT-3 and ChatGPT. Balaji reflects on how these models can encode complex concepts and navigate abstract ideas, surprising many with their capabilities in language understanding and generation, which were previously underestimated.
"evolution that deals with kind of a much denser space that actually does a pretty good job with linear interpolation? you the problems the the problems that actually works very well with linear interp..."
The Challenge of Spatial Reasoning
Balaji discusses the limitations of AI in spatial reasoning compared to its proficiency in language tasks. He highlights the complexities involved in teaching AI to navigate and understand physical spaces, emphasizing the challenges that arise from the differences in human and AI cognitive evolution.
"Yeah. Like before the Chad GPT moment, it wasn't obvious to me that just then afterwards you're like, okay, language is sophisticated enough to encode almost any concept about the world, right? or or ..."
Understanding AI's Self-Replication Constraints
In this segment, Balaji addresses the current constraints on AI regarding self-replication and autonomy. He explains that AI lacks the embodiment and goal-setting capabilities necessary for independent action, which alleviates fears about AI surpassing human control in the near future.
"yeah the same thing but but but but from the world to the world model was the human and then everything else and then the lang agree. Yeah, that's right. But but it was it was a little surprising to m..."
The Importance of Prompting in AI
Balaji emphasizes the critical role of prompting in AI functionality, likening it to navigating a high-dimensional space. He discusses the complexities involved in crafting effective prompts and how they serve as essential tools for guiding AI behavior and responses.
"actually talk about one thing you did say which is self-replication right I don't think of that as a forever constraint on AI I think of that as a today constraint and the reason I think of that as a ..."
Closing the Control Loop in AI
This segment delves into the challenges of closing the control loop in AI systems. Balaji explains the necessity for AI to understand its own limitations and the implications of producing outputs that may be out of distribution, highlighting the complexities of ensuring reliable AI behavior.
"arose to help your survival replication to help stand outside of yourself to to be able to see the That's right. That's right. So, right now AI does not really have goal setting. It doesn't have repro..."
Self-Reflection and AI's Knowledge Boundaries
Balaji discusses the concept of self-reflection in AI, stressing the importance of models recognizing their own knowledge boundaries. He explores how this self-awareness is crucial for improving AI's performance and ensuring it does not produce nonsensical outputs.
"in, right? And so a prompt is a very highdimensional direction vector even if you account for the fact that many potential prompts of just strings of random characters wouldn't wouldn't be interesting..."
The Complexity of AI Prompts as Programs
In this segment, Balaji articulates the idea that prompts function as tiny programs within AI systems. He contrasts the nature of prompts with traditional APIs, emphasizing the need for a rich vocabulary and subject knowledge to effectively utilize AI's capabilities.
"So it could it could produce in fact it's optimized to fake it. Yes. And so if you could if you could tell it if you could say hey listen produce a bunch of directions by feeding the last direction in..."
The Age of the Phrase: Unlocking AI Potential
Balaji concludes by discussing the significance of concise phrases in the current AI landscape. He highlights how the ability to articulate ideas succinctly can unlock the full potential of AI, drawing parallels to social media and crypto, where brevity and clarity are paramount.
"Yeah. So, it's like real-time events, obscure or niche academic fields, um, and specialized subfields behind pay walls, local and regional information, human emotion, intent or experience, um, private..."
AI as Superhuman Intelligence
In this segment, Balaji explores the analogy of AI as a polytheistic pantheon of superhuman intelligences, drawing parallels with Hindu and Norse mythology. He discusses the interpretability of AI systems and the potential for understanding their inner workings, suggesting that as we gain insights into AI, the perception of these systems as 'gods' will diminish. This conversation delves into the philosophical implications of AI's evolving role in society.
"phrase, which is the prompt, the 140 character tweet, and the 12 words for your crypto password, right? these these phrases of power in AI, in social media, and in crypto just unlock everything. So th..."
AI vs. Crypto: The Grounding Problem
Balaji contrasts the probabilistic nature of AI with the deterministic principles of cryptocurrency. He argues that while AI can generate content, it struggles with grounding in reality, which crypto can help establish through verifiable data. This segment discusses the limitations of AI in providing true grounding and the potential for blockchain technology to enhance the reliability of information.
"Well, that's right. I mean, that's the thing is actually what's interesting is that the um the interpretability work that anthropic and others have done and the work on like groing or what have you, r..."
On-Chain Data and AI Citations
This segment focuses on the integration of on-chain data with AI outputs. Balaji explains how AI can reference on-chain data for financial and social assertions, enhancing the credibility of its responses. He discusses the challenges of grounding AI in the physical world and the importance of ensuring that data entering AI systems is accurate and trustworthy.
"And so, that's, you know, like like the hard barriers, right? I mean, I generally agree. I don't I mean I don't think crypto solves the grounding problem, right? I mean it's a it's it's a mechanism yo..."
Capturing Reality with Crypto Instruments
Balaji introduces the concept of crypto instruments as a means to capture and verify data in real-time. He discusses the potential for using cryptographic methods to ensure the authenticity of scientific data and other forms of information. This segment highlights the intersection of technology and verification, emphasizing the need for reliable data in an increasingly digital world.
"world grounding. Like I say something, I am a human being. I, you know, you have no idea what I said is true or not true. You know, there's there's a geographic place where there's a picture of the ge..."
Visual vs. Verbal: AI's Strengths and Weaknesses
In this segment, Balaji contrasts AI's effectiveness in visual tasks versus verbal reasoning. He explains that AI excels in generating images and visual content due to its stateless nature, while it struggles with complex verbal tasks that require deeper verification. This discussion sheds light on the inherent limitations of AI in understanding and processing language compared to visual data.
"It's just a data injust problem is a long-standing problem in computer science and over time everything you're saying is going to be more and more true because over time we're going to be more and mor..."
The Challenges of AI in Markets and Politics
Balaji discusses the limitations of AI in dynamic and adversarial environments such as markets and politics. He argues that AI struggles with time-varying and rule-varying systems, where human intuition and adaptability are crucial. This segment emphasizes the need for human oversight in AI applications within complex systems, highlighting the challenges of relying solely on AI for decision-making in these fields.
"you're doing the easier it is to and now the interesting concept is how much go one more thing which is like I for for me again like I spend most of my time in like software and engineering the big di..."
AI's Role in Prompting and Verification
In this segment, Srinivasan explores the commercial implications of AI prompting and verification. He argues that as AI-generated outputs become more prevalent, businesses will increasingly invest in systems to verify and proctor these outputs. This shift reflects a growing trend towards a low-trust society, where verification becomes essential to ensure the reliability of AI-generated content.
"is I wasn't even thinking of the stock market as complex different but you're right on the subset of Mandel wrote this great book on the fact that these things are are these cha they're super chaotic ..."
Amplified Intelligence: The New CEO
Balaji discusses the concept of 'amplified intelligence,' suggesting that AI enhances human capabilities rather than replacing them. He notes that skilled individuals can leverage AI to improve productivity and decision-making, effectively becoming 'CEOs' in their own right. This segment emphasizes the importance of communication and clarity in utilizing AI tools to maximize their potential.
"thing and that that maps to KYC that maps to like in a bad way, you know, the the glass cases in Walmart, right? In a sense, a low trust society is spending more and more and more on verification and ..."
AI's Competitive Landscape: The Job of Previous AIs
Srinivasan argues that AI does not take jobs from humans but rather replaces previous AI systems. He explains how each new AI iteration competes with its predecessors, creating a dynamic landscape where AI tools continuously evolve. This segment highlights the complementary relationship between humans and AI, where new AI technologies enhance existing workflows rather than displacing human roles.
"ways to explain trade-offs constrained more constrained. Yeah. Basically constrained languages that reduce ambiguity. This is literally is strictly an efficiency thing, right? And so like someone that..."
The Polytheistic View of AI Intelligence
In this thought-provoking segment, Balaji and his co-hosts discuss the polytheistic view of AI, where multiple AI systems coexist and learn from each other. They explore the idea that while AIs may appear similar, they can have distinct capabilities shaped by their training data and design. This conversation delves into the implications of AI convergence and the potential for a core intelligence that underpins various AI models.
"Let me know your thoughts. Can I can I actually this is an adjacency to what you're just saying but can I actually push on something you said previously because I actually agree with your polytheistic..."
The Complexity of AI Training and Specialization
Srinivasan addresses the trade-offs involved in training AI models for specific tasks. He explains that while specialization can enhance performance in one area, it may detract from capabilities in others. This segment underscores the need for a plurality of AI models to address diverse challenges, reflecting the complex nature of AI development and deployment.
"forth there's and but then some people have you know much better vision or they have much better speech or something like that right so there may be some distilled kind of thing oh by the way another ..."
AI for Experts vs. Non-Experts
In this segment, the discussion shifts to the different ways experts and non-experts interact with AI. Balaji highlights that while non-experts can use AI to achieve satisfactory results, experts can leverage their domain knowledge to extract superior outcomes. This distinction emphasizes the ongoing relevance of human expertise in an increasingly AI-driven landscape.
"the best, we see this a lot, right? Which is uh uh you know, a model that's very very good for certain parts of code is just not going to be generally good at other things because those are the trade-..."
The Versatility of AI in Various Domains
Balaji concludes by discussing the versatility of AI across different domains. He notes that while AI can perform a wide range of tasks, its effectiveness often depends on the interface and context in which it is used. This segment reflects on the potential for AI to enhance productivity in both casual and professional settings, highlighting the importance of user experience in AI applications.
"specialist model I would have to become a specialist would be the argument to our previous one so I think I think listen for casual use I and use these models for whatever I want, but like to really u..."
The Versatility of AI Models
Balaji explores the versatility of AI models, comparing their capabilities in logical and probabilistic reasoning. He discusses how AI has evolved to excel in both deterministic tasks and probabilistic text generation, suggesting a potential future where these two approaches could be merged. This segment delves into the implications of such advancements for AI's role in various applications.
"But thus far, these are kind of very different user bases, right? There's professional coding versus basically casual coding. I mean, you know, part of it is which is interesting and a little counteri..."
The Trade-offs of AI Systems
This segment addresses the fundamental trade-offs in AI system design, particularly between determinism and flexibility. Balaji argues that while hybrid systems could theoretically exist, the complexity of the universe may prevent a single AI from effectively managing both deterministic and probabilistic tasks. The conversation raises questions about the future of AI and its integration with traditional software.
"Yeah. This this very old school, you know, systems part of me thinks that there's a fundamental trade-off here, which is like you can trade off what's that? Well, I I feel like you can trade off you c..."
Exposing AI Internals for Better Understanding
Balaji proposes the idea of exposing the internal workings of AI systems to enhance user understanding and trust. He suggests that visualizing AI outputs, such as through spectrograms for audio, could help users gauge AI performance. This segment emphasizes the importance of transparency in AI and how it can improve user interaction and confidence in technology.
"basically but humans can do it. No, we don't. We use calculators and we use software. Like the whole reason we built the whole I know, but software is because humans can't do I sure but at some level ..."
Killer AI: Drones and Digital Borders
In this segment, the conversation shifts to the implications of AI in warfare, particularly through the use of drones. Balaji discusses how countries are developing autonomous drones that can operate without direct control, raising concerns about digital borders and security. He highlights the potential for AI to redefine geopolitical boundaries and the nature of warfare in the digital age.
"like for example we could push on here that you know we're very early innings of yeah like like colored text for example in terms of its level of confidence you know stuff like that like yellow red gr..."
The Surveillance State and AI
Balaji shares a personal anecdote about surveillance in China, illustrating the extent to which conversations can be monitored. He discusses how AI can change the balance of power in surveillance and control, emphasizing the need for awareness of these technologies' implications on privacy and security. This segment underscores the intersection of AI, governance, and individual freedoms.
"the alternative to that of having quote defensible borders is basically an encrypted state where you don't even know where it is on on the face of the earth. And what I mean by that is can you make a ..."
AI's Role in Information Control
Balaji discusses the concept of Total Information Awareness and how AI enhances the ability to track individuals through data. He argues that AI's capacity to process vast amounts of information changes the landscape of surveillance and control, raising concerns about privacy and autonomy.
"is the mountains are high and the emperor is far away, right? And China always had a different conception of the balance of power between the government and the people than the west did. Sure. On the ..."
The Future of Power and Cryptography
This segment delves into the future of power dynamics influenced by AI and the importance of cryptography as a countermeasure. Balaji suggests that as AI capabilities grow, individuals will need to find ways to protect their autonomy and privacy from state surveillance.
"Yeah. Yeah. They can they can synthesize you know there was something maybe you know maybe you know this thing Martin I think it was called TIA total information awareness in Iraq at a certain point w..."
The Anti-AI Backlash
Balaji addresses the emerging anti-AI backlash, drawing parallels to the anti-crypto sentiment. He discusses how various sectors, including media, are reacting to AI competition and the potential consequences of resisting technological advancement.
"there. So some thoughts on balance of power since you talked about that. Okay, last one. I think that there's going to be there already is an anti- AI backlash that's like the anti-crypto backlash and..."
Global Wage Disparities and AI
In this segment, Balaji examines the impact of AI on global labor markets, highlighting the wage disparities between developed and developing countries. He discusses how AI could elevate wages abroad while potentially lowering them in the West, leading to significant economic shifts.
"where you know in some on some like artist forums or whatever they'll say, are you an AI supporter? Have you heard that? You know, you know that you know like AI artists spend as much time on building..."
AI as a Political Tool
Balaji concludes with insights on how AI serves as a powerful political tool, capable of mobilizing public sentiment across the political spectrum. He reflects on the historical context of technology's role in society and its potential to shape future political landscapes.
"To your previous point, I just think this is so important. Like I agree there's going to be a huge backlash and I think some of it's going to be rooted in like the experience of individual people like..."