Humans often overlook the workings of various technologies that have become integral to our daily lives, such as crypto arbitrage and bots, social media bot armies, algorithmic feeds, generative art, AI/ML illustrations, and background automation. Many of us are so deeply immersed in using software that we can be considered software cyborgs. Today, we aim to explore how these trends are already shaping the landscape, specifically delving into the influence of AI and DePIN.
Navigating the On-Chain/Off-Chain Divide in Web3
There's a clear distinction between on-chain and off-chain activities in the crypto space. Money-related processes, like dollar bank deposits, occur off-chain, while things like DAI exist on-chain. Similarly, real-world assets in tokenized form are off-chain, whereas wrapped tokens or liquid staking are on-chain. Traditional documents like passports are off-chain, while digital items like POAPs and NFTs are on-chain. The same concepts, be it money, financial instruments, or identity, can be represented in different spaces.
This distinction applies to artificial intelligence as well. AI can operate off-chain and occasionally interact with Web3 for specific actions. In such cases, we require services that function like oracles or on/off ramps for machine intelligence. Just as fintech created embedded financial products, there's a potential for GPTs (like ChatGPT) to navigate crypto markets with the support of API software services. The challenge then becomes ensuring the honesty of AI and verifying its outputs within trustless environments.
Integrating crypto custody and control into AI agents is anticipated to become a key value proposition in Web3. Some teams are exploring incorporating crypto technologies, such as zero-knowledge proofs, into the computational processes of machine learning models. This would provide proof that a particular AI system performed as intended, preventing deception similar to Bernie Madoff with a spreadsheet. While this approach is still in its early stages, the value of verifying interactions with AI is apparent.
Other teams are considering shifting the entire large language model (LLM) and neural network stack to a decentralized infrastructure. Given the open-source nature of much of the generative AI movement, deploying and maintaining open-source models on decentralized protocols is conceptually possible. This could distribute the computational load and create incentives for high-quality machine intelligence services.
However, despite some teams exploring these strategies, the practical implementation is still largely in science fiction. Even centralized AI teams grapple with understanding demand and achieving profitable unit economics. Splintering the provision of such services into protocols and decentralized autonomous organizations (DAOs) at this stage is considered early in the development of this space.
Decentralized Physical Infrastructure (DePIN) Unveiled
Digging into the foundational layers, we encounter DePIN, which stands for decentralized physical infrastructure. The basic DePINs utilize coin-protocol incentives, resembling the proof-of-work mining in Bitcoin. Nodes contributing storage, compute power, GPU, or training data receive rewards for their work. However, we anticipate a shift from this coin-based model to a dApp running on computational blockchains like Peaq, Solana, or an EVM rollup, similar to the evolution seen in projects like Helium.
While Web3 previously relied on centralized cloud services for decentralized networks, the practicality of this approach for decentralized AI is uncertain due to cost and demand factors. Centralized AI providers demonstrate superior efficiency and organization, making them more formidable than protocol-centric competitors. Therefore, scalable DePINs could be a critical resource unlock for on-chain crypto-AI.
What makes DePIN intriguing is its role in onboarding machines into Web3. Machines will likely require decentralized finance (DeFi) and access to intelligence-as-a-service. Picture a future where fleets of self-driving cars download the latest models from DePIN AI networks managed by DAOs and incentivized by tokens—an idea already undergoing experimental phases.
Frameworks for connecting, standardizing, and managing populations of AI agents with diverse objectives are emerging. Autonolas, for instance, generates bots for trading and governance. Notable experiments include Botto, a generative AI model curating art for a DAO, and Numerai, a hedge fund running a token-incentivized competition for AI algorithms.
An important distinction lies in the scope of AI services. In one scenario, AI enhances specific features, like a digital wallet engaging in conversations about held tokens. However, no substantial industry transformation occurs. Alternatively, envision a world where OpenAI acts as a new iOS, and its GPT store becomes the app store for AI applications. Here, AI becomes a platform addressing general problems, potentially subsuming crypto as one of its functionalities.
In this latter scenario, concerns arise about centralized custody of AI agents, posing risks across data, privacy, and finance. Custodial arrangements often lead to principal-agent issues, necessitating legal and regulatory frameworks to enforce fiduciary duty. In big tech, the inevitability of AI regulation to protect individuals seems apparent.
Web3's Self-Custody Revolution for AI Agents
To counter the potential risks associated with centralized custody of information and AI agents, Web3 introduces the concept of self-custody. Imagine generating advanced AI models, such as GPTs, on a centralized platform but having the ability to transfer those trained models into a crypto wallet for individual ownership. As the landscape evolves with successful open-source models functioning well on decentralized infrastructure, the core value proposition of Web3 is expected to be the custody and control of AI agents through crypto mechanisms.
Current projects aiming to verify the authenticity of digital media in the era of AI align closely with this concept. The emphasis is on ensuring that digital content is real and resistant to manipulation.
Consider NFT minters linked to image generation or large language model (LLM) engines. This connection aligns machine-generated creations with the market opportunities presented by Web3 decentralized finance (DeFi). However, the commercial traction around such digital assets remains relatively low, influenced by factors such as the state of NFT markets, the quality of machine-generated content, and the perceived utility of these digital assets.
In any case, this evolving landscape presents an incredibly intriguing design space for entrepreneurs. Since the inception of Generative Ventures, we continue to be amazed by technologist's diverse creativity and enthusiasm as they explore the realms of the possible and pave the way forward.
Conclusion
In conclusion, the intersection of AI and Web3 technologies has the potential to revolutionize the way we interact with and use technology. The concept of decentralized AI, made possible by DePIN, offers a promising solution to the challenges of centralization and provides a foundation for the next generation of innovators to build upon. The integration of AI and Web3 also raises important questions about the ownership and custody of AI agents, highlighting the need for responsible and ethical practices in this rapidly developing field.
As we move forward, it is essential to consider the implications of these technologies on society and ensure that their benefits are accessible to all. By fostering collaboration and innovation, we can harness the full potential of AI and Web3 to create a brighter, more equitable future.