Illia Polosukhin, co-founder of NEAR and co-author of the 'transformers' paper, shares insights on the revolutionary convergence of AI and crypto. He explores the concept of user-owned AI and its potential to replace traditional apps, emphasizing decentralized control. The conversation also delves into AI's evolution, from fascination to functionality, and its balance with human interactions. Moreover, he discusses the geopolitical implications of AI sovereignty and NEAR's innovative approach to scalability in blockchain technology.
The convergence of AI and cryptocurrency raises concerns about centralization, necessitating user-owned AI to enhance data sovereignty and privacy.
Innovative monetization strategies are crucial for AI models to prioritize user-oriented outcomes while navigating a competitive landscape dominated by profit motives.
The future of technology may see a shift from traditional apps to agent-driven AI interactions that streamline user experiences across digital platforms.
Deep dives
The Intersection of AI and Crypto
The discussion revolves around the convergence of artificial intelligence (AI) and cryptocurrency, highlighting that both fields face similar issues regarding control and centralization. The current dominance of a few corporations in AI parallels the monopolistic tendencies seen in the early days of the internet. Companies that dominate AI have significant control over user data, raising concerns about privacy and potential biases in AI systems. The need for a decentralized approach, facilitated by blockchain technology, presents a viable solution to these challenges by providing user-owned AI that can operate independently of corporate control.
Transforming User Ownership through Blockchain
The concept of user-owned AI is further explored, emphasizing the necessity for blockchain to ensure data sovereignty. A potential model is proposed where AI can be developed and operated without direct access to user data, thereby enhancing privacy. Technologies such as zero-knowledge proofs are identified as tools that could support this form of confidential computation. This approach not only addresses issues of data privacy but also creates a monetizable framework for AI development that aligns with user interests.
Challenges in AI Monetization
The episode underscores the economic challenges that AI models face in a competitive landscape where monetization is essential. The conversation points out that while many companies race to develop AI technologies, a focus solely on profit could compromise user-oriented outcomes. Models driven by advertisement biases could lead to undesirable results, highlighting the importance of prioritizing user success over corporate revenues. Addressing this challenge requires innovative monetization strategies that create fair conditions for both users and developers.
The discussion introduces advancements in hardware, specifically mentioning Intel and NVIDIA’s technologies that enable confidential computing. These developments allow for secure data processing where even the hardware providers cannot access user data, thus enhancing privacy and security. This confidential computing framework supports a decentralized approach to AI, facilitating private model deployment without compromising user data. This innovation paves the way for the creation of an AI ecosystem that respects user privacy while enabling effective computational tasks.
Future of Interactions with AI Agents
As the conversation delves into the future of AI interactions, it posits that traditional websites and applications may transition toward more agent-driven models. Users will interface with AI agents that autonomously navigate websites and execute tasks, leading to a streamlined and efficient experience. The role of AI as facilitators in everyday decision-making is emphasized, indicating that these agents can manage various transactional interactions, thus simplifying user engagement with complex systems. This shift signals a paradigm change in how users will relate to technology, prioritizing engagement through personal AI assistants.
Welcome to web3 with a16z. I’m your host, Robert Hackett.
In this episode, we're diving deep into one of the most intriguing intersections in tech today: AI and crypto.
To help us unpack it, we're joined by Illia Polosukhin — co-founder of the crypto protocol NEAR and co-author of the groundbreaking 2017 "transformers" paper that kicked off the current AI boom. Ilia has been early to some of the biggest recent tech trends, and today he brings us a rare, panoramic view of the tech industry’s cutting edge.
Together we explore what the phrase “user-owned AI” really means; why the so-called agentic internet — that is, a world where your AI assistant talks directly to services on your behalf — might replace the very notion of websites and apps as we know them; and much more.
Timestamps:
(0:00) Introduction (3:40) Centralization and Challenges of AI (6:17) "User-Owned" AI (12:14) Confidential Computing and AI (17:51) The Birth of Transformers (22:33) NEAR AI and Crowdsourcing (27:56) AI Agents and Future Applications (31:04) The End of Websites and Applications (34:08) Dead Internet Theory & Distinguishing Humans (41:49) Open Source vs. Open Weight Models (43:48) Geopolitical Implications of AI (46:55) NEAR Protocol and Blockchain Scaling (59:29) The Role of Humans in an AI World
Resources:
Attention is all you need by Vaswani et al. (Conference on Neural Information Processing Systems 2017)
As a reminder, none of the content should be taken as investment, business, legal, or tax advice; please see a16z.com/disclosures for more important information, including a link to a list of our investments.
Remember Everything You Learn from Podcasts
Save insights instantly, chat with episodes, and build lasting knowledge - all powered by AI.
