A Positive Vision for the Future: Part 2 with Illia Polosukhin of NEAR
Timestamps are as accurate as they can be but may be slightly off. We encourage you to listen to the full context.
Podcast Summary
In part two of his conversation with Nathan Labenz, Illia Polosukhin, co-author of "Attention Is All You Need" and founder of NEAR Protocol, explores his ambitious vision for an AI-powered future. The discussion covers how AI coding assistance is already transforming software development by enabling personalized automation for everyone, moving beyond traditional software-as-a-service models. (03:33) Polosukhin envisions a world where AI agents create a unified intelligence layer across all devices, eliminating traditional user interfaces in favor of natural language interactions that predict and fulfill our needs proactively. The conversation delves into how this transformation will reshape markets by enabling direct connections between buyers and sellers, potentially reducing the role of middlemen and advertising. (33:03)
- Main theme: Polosukhin presents a comprehensive vision for how AI will transform computing, economics, governance, and society, emphasizing the importance of user-owned AI systems and decentralized governance models over centralized control by tech giants.
Speakers
Illia Polosukhin
Illia Polosukhin is the co-author of the groundbreaking "Attention Is All You Need" paper that introduced the Transformer architecture, fundamentally changing how AI systems process language. He is also the founder of NEAR Protocol, a blockchain platform designed to support AI applications and enable decentralized ownership of artificial intelligence systems. Previously, he worked at Google Research where he contributed to foundational AI research before transitioning to blockchain technology to solve the challenge of fairly compensating data contributors for AI training.
Nathan Labenz
Nathan Labenz is the host of The Cognitive Revolution podcast and a technology entrepreneur focused on AI applications. He has extensive experience testing and implementing AI systems, including early access to frontier models like GPT-4, and regularly engages with leading researchers and builders in the AI space.
Key Takeaways
Personal AI Will Replace Traditional Software Interfaces
Polosukhin envisions a fundamental shift from traditional software-as-a-service models to personalized AI systems that can code solutions on demand. (03:33) Rather than learning complex software like Salesforce, users will simply describe their needs in natural language, and AI will create custom automation. This eliminates the need for one-size-fits-all solutions and removes the complexity of traditional user interfaces. For example, instead of hiring someone to configure Salesforce and build Telegram integrations, a user can simply ask their AI to create a custom CRM with built-in Telegram functionality. This represents a shift from software as a craft requiring technical expertise to problem-solving through natural language communication with computers.
AI Agents Will Transform Economic Markets Through Direct Connections
The current consumer economy built on advertising and middlemen will be disrupted as AI agents enable direct relationships between buyers and sellers. (38:03) Polosukhin explains that AI can eliminate layers of distribution by connecting consumers directly with manufacturers and farmers. Instead of shopping at Costco, your AI could purchase directly from suppliers, with other AIs handling logistics and batching for efficiency. This could reduce waste significantly - currently 40% of food in the US is thrown away due to over-provisioning by stores that don't know actual demand. When AIs can predict and coordinate purchases in real-time, supply chains become far more efficient and responsive to actual consumer needs.
Status Games Will Replace Economic Competition in an AI-Abundant World
As AI handles most productive work, human competition will shift toward status games and niche communities focused on specialized interests rather than economic productivity. (49:00) Polosukhin argues that people will find meaning and compete for recognition in athletics, video games, collecting unique artifacts, and other pursuits that don't directly contribute to GDP. This mirrors existing examples like professional athletes who may be less wealthy than billionaires but command significant respect and attention. He predicts a proliferation of small communities where people can achieve status and recognition within specific niches, moving away from money as the primary status indicator toward more diverse forms of social competition and achievement.
Formal Verification Is Essential for Blockchain Security in an AI World
To enable safe AI-generated blockchain applications, NEAR is working toward formal verification systems that can mathematically prove the correctness of smart contracts. (25:41) Polosukhin explains that blockchain's adversarial, monetary environment requires stronger guarantees than traditional software. The goal is for users to specify high-level intentions (like "don't lose my money") and have the system prove these properties will hold before executing transactions. This involves chaining proofs through complex interactions - if money is lent to someone, the system must prove either repayment or successful liquidation of collateral. This approach extends beyond simple low-level code correctness to verify that applications actually do what users intend them to do.
AI Governance Can Enable True Representative Democracy
Traditional governance systems compress complex decisions into infrequent votes for representatives, but AI delegates could enable continuous, detailed democratic participation. (58:33) NEAR is experimenting with AI delegates that vote on behalf of token holders based on their stated values and decision-making frameworks. Users can inspect the AI's reasoning process, test its alignment with their views, or create alternative delegates. This could eventually scale to every individual having their own AI that participates in governance on their behalf, eliminating the principal-agent problem where representatives pursue their own interests rather than their constituents'. Polosukhin envisions this extending to traditional government, potentially enabling 300 million Americans to have meaningful input on every policy decision through their AI representatives.
Statistics & Facts
- Currently in the US, 40% of all food is thrown away due to over-provisioning by stores because they don't know exactly what people will buy. (43:43) Polosukhin cites this as an example of inefficiency in current supply chains that AI could eliminate through better demand prediction.
- Right now in US manufacturing, workforce churn is 300% annually, meaning companies need to hire three people for every one job each year because employees keep quitting. (18:08) This makes robotic automation economically attractive even beyond just cost considerations.
- No specific third statistic was provided in this episode that meets the criteria for inclusion.
