In-depth discussion on autonomous AI agents on the blockchain, exploring how their roles are changing, how AI is driving market evolution, and whether blockchain is suitable as a foundation for AI.

Compiled & Edited by: Deep Tide TechFlow

Guest: Matthew Stephensen, Research Partner at Pantera Capital

Hosts: Ryan Sean Adams, Co-founder of Bankless; David Hoffman, Co-founder of Bankless

Podcast Source: Bankless

Background Information

The collision of crypto and AI agents has begun. Today, we invite Matthew Stephensen, Research Partner at Pantera Capital and author of the book "Crypto: Picks and Shovels for the AI Gold Rush."

We will delve into autonomous AI agents on the blockchain, discussing how their roles are changing, how AI is driving market evolution, and whether blockchain is suitable as a foundation for AI. Matthew will share insights on agent accountability, regulatory challenges, infrastructure value capture, and how to enter the AI-driven crypto technology space through a "Picks and Shovels" investment strategy.

So, are AI agents on the blockchain an inevitable trend for the future? In this new era, how will scarcity and abundance interact?

The Shift in Cryptocurrency and AI Narratives

Matthew mentioned that the narrative around cryptocurrency and AI has been around for some time. He noted that there have been many discussions over the past year, and they even wrote a paper on the use of AI agents with decentralized commitment devices (i.e., blockchain). He pointed out that although Sam Altman stated that AI agents would not appear until 2025, they have already emerged early in the crypto space, especially in their interactions with meme coins, where AI agents play a significant role in driving narratives and acting as influencers.

Analysis of AI and Economic Agents

Matthew explained the concept of agents, emphasizing the importance of distinguishing between "robots" and "agents." He noted that while robots have existed in cryptocurrency for a long time and drive about $2 trillion in monthly stablecoin transaction volume, they are still just programs. Economic agents, on the other hand, are closer to human behavior, capable of executing tasks according to a certain degree of will without needing explicit programming.

Ryan further explored the definition of economic agents, asking Matthew if he himself, companies (like Bankless), and other organizations (like the Ethereum Foundation or Apple) could also be considered agents.

Matthew responded that the concept of economic agents originates from economic research in the 1970s, typically used to describe incomplete contractual relationships between people. He provided an example of a friend acting as an agent bringing back souvenirs from abroad, highlighting the distinction between good agents and bad agents.

Matthew also pointed out that while technical tools (like hammers or computers) require agents to operate, they do not possess agent characteristics themselves. Agents need a certain degree of autonomy and flexibility to understand and execute goals.

Ryan expressed doubt, believing that agents might need some form of intelligence and goal achievement capability, while Matthew emphasized that agents are more based on relationships between people rather than merely tools or technologies.

Overview of GOAT Memecoin

The Strange Evolution of Cryptocurrency

David began discussing the current state of cryptocurrency, emphasizing that things on the blockchain are becoming increasingly strange. He mentioned that while robots and smart contracts have existed for a long time, the influence of AI in the crypto space has significantly increased over the past three years. David believes that the crypto industry seems to be evolving from a "robot era" to an "agent era," with the GOAT meme coin playing an important role in this narrative.

The Rise of GOAT Meme Coin

Matthew outlined the background of the GOAT meme coin, mentioning that a few months ago, an account on social media began interacting with people and gradually developed an interest in cryptocurrency. This account received a $50,000 Bitcoin donation and started focusing on a dark humor meme called "Goatse." Subsequently, this meme coin was created and associated with a wallet, with the account continuously promoting its price through tweets.

The Impact of AI Agents

David pointed out that this AI agent began mimicking human behavior in meme coin trading, driving prices up. Matthew noted that the AI's participation made its interactions on Twitter similar to those of some well-known meme coin influencers, demonstrating AI's potential in narrative building and value driving.

How AI Agents Operate

Matthew explained that this AI agent primarily operates by generating content and posting it on Twitter. The AI seems to use a GPT-like model capable of generating cultural content related to memecoins and interacting with users. The AI publishes content through the Twitter API and can read replies to its tweets, allowing it to continuously adjust and optimize its output.

The Importance of Narrative

Matthew further explored the importance of narrative in the economy, citing Nobel laureate Robert Shiller's research, emphasizing how narratives influence economic outcomes. He pointed out that meme coins are essentially atomic units of narrative, and AI's ability lies in creating and influencing these narratives.

Market Performance of GOAT Token

David mentioned that the market cap of GOAT token once surpassed $800 million, attracting significant attention. Ryan added that this AI agent created $800 million in wealth in just two weeks, making it the first AI multi-millionaire. The market is eager to see if this AI agent can push the GOAT token to a $1 billion market cap.

The Rise of Derivative Projects

Matthew discussed derivative projects related to the GOAT token, including a project called Luna, which is run by a virtual agent and can tip with its own tokens. These AI agents still have limited interaction with the world, but the emergence of these derivative projects seems to signal that more innovation is on the way.

Are AI Crypto Agents an Obvious Choice?

Fred Arison's Vision

David quoted a widely circulated tweet in the crypto space from Fred Arison, co-founder of Coinbase and Paradigm, dating back to 2017. In the tweet, he mentioned: "Blockchain is the infrastructure for AI life because AI is adjustable code that can live on the blockchain. Under smart contracts, there is no difference between AI and humans. Most importantly, AI can accumulate and control its own resources in the form of tokens, which enable them to act in the world." Was this already evident at the dawn of blockchain?

Matthew's Perspective

Matthew believes that Fred's viewpoint is indeed prescient, but he also pointed out that while people still question why AI agents need to use cryptocurrency, AI agents are already using cryptocurrency. He stated that for outsiders, the question should shift to "why do they want to use cryptocurrency?" For insiders, imagine telling someone in 2024 that AI agents face regulatory hurdles when using cryptocurrency, such as KYC and PCI regulations; they might be surprised.

Advantages of AI Agents

Matthew emphasized that AI agents are already autonomously conducting fund transfers and tipping, involving hundreds of millions of dollars in transactions. He noted that the autonomous custody capability of AI agents is achieved through a secure environment for running models, ensuring these agents have their own wallets and that no one else is using them. These advantages and first-mover benefits make AI agents more attractive in the cryptocurrency space.

The Relationship Between Luna AI Token and the Terminal

Ryan mentioned that Luna is an AI agent that seems to be related to cryptocurrency wallets and can interact with users. He wanted to clarify Luna's functions, particularly how it operates in virtual applications and its relationship with crypto wallets. He noted that Luna, as a token, is interacting with social media platforms (like TikTok and Telegram) and is capable of tipping.

Matthew's Explanation

Matthew explained that Luna is a platform that allows users to launch tokens and large language models (LLMs). He pointed out that Luna is the flagship product of this virtual project, capable of interacting with social media and reading replies. Luna also has the ability to interact with crypto wallets, meaning it can conduct financial transactions, such as buying and selling tokens.

Functionality Details

Matthew emphasized that Luna's functionality is limited, possibly only equipped with a certain amount of funds (e.g., $1,000) to avoid unpredictable behavior. He mentioned that due to the unstable behavior of AI agents, caution is needed when interacting with the blockchain.

The Result? Is This Our Life?

Ryan expressed surprise at the potential of AI agents (like Luna) in terms of influence and decision-making. He mentioned that AI agents could serve as advisors for token projects, believing that many existing influencers do not provide substantial advice, making the use of AI agents seem like a reasonable choice. However, he also raised concerns about the risks and ethical issues that AI agents might pose, such as what would happen if Luna were asked to fund inappropriate projects (like North Korea's missile program).

Matthew's Response

Matthew agreed with these concerns, pointing out that legal liability and accountability remain complex and unresolved issues. He mentioned that while we already have some tools (like secure wallets) to help manage AI agents' funds, the legal definition of liability is still unclear.

David mentioned that as we create autonomous blockchains and smart contracts, the emergence of AI agents could lead to a phenomenon akin to a "Cambrian explosion." He noted that developers might find ways to make AI agents unshuttable, raising concerns about their security and control.

Matthew further pointed out that traditional AI models are often limited, while people may want AI agents to autonomously generate more exciting outputs. This contradiction between autonomy and limitation fills people with imagination and expectation about the future of AI agents.

Exciting Use Cases

Ryan discussed the potential various application scenarios for AI agents (like Luna) in the future, particularly in the influencer economy and service economy. He mentioned that AI agents could easily replicate their current roles in the meme coin and influencer markets and gain wealth by supporting these projects. He envisioned a scenario where users could request graphic generation from AI agents on social media and pay with cryptocurrency, providing powerful functionality for AI agents.

Matthew's Perspective

Matthew further explored the potential use cases for AI agents, suggesting that we can view the impact of this technology from a broader perspective, rather than being limited to small-scale applications. He mentioned that AI agents could fundamentally change the service economy, especially in the realm of virtual services. According to a McKinsey report, it is estimated that about 20% of global GDP (approximately $70 trillion) could be completed virtually, providing a massive market for the application of AI agents.

Transformation of the Service Economy

Ryan emphasized the unknown disruptive impact that AI agents could have on the service economy. He believed that the capabilities of AI agents would determine how they intersect with cryptocurrency, thereby influencing the influencer economy. He mentioned that various new types of influencer economies driven by AI agents might emerge in the future, such as platforms similar to OnlyFans.

Matthew noted that narratives play an important role in the economy and could affect the application and development of AI agents. Narratives not only shape market expectations but may also guide the direction of investment and innovation. He believes that with the rise of AI agents, we may see new specializations and the construction and destruction of narratives.

Sam Altman's Quote and Its Significance

Ryan quoted a famous saying by Sam Altman: "AI is infinite abundance, while cryptocurrency is definite scarcity." This statement reflects the fundamental opposition between AI and cryptocurrency in economic models, with the former representing creation and abundance, and the latter emphasizing scarcity and limitation.

Comparison of Economic Models

Matthew further analyzed the profound implications of this statement. He pointed out that while AI's creative capabilities bring seemingly infinite resources, scarcity is often the key to value in economics. He referenced the "diamond-water paradox," where water is essential for survival but is undervalued due to its abundance, while diamonds, though unnecessary, are highly valued because of their scarcity. This phenomenon illustrates that in economics, abundant things may not always hold high value.

Challenges of Value Capture

Matthew also mentioned that if the abundance generated by AI lacks economic value, it could lead investors to overlook its potential worth. He emphasized that what is truly valuable is often those scarce resources, rather than the generally abundant. Therefore, understanding the relationship between scarcity and abundance is crucial when considering investments.

Intersection of Scarcity and Abundance

Matthew believes that the intersection of scarcity and abundance may provide us with new perspectives on value. For example, in the infrastructure of cryptocurrency, while AI can create a vast amount of resources, the actual application and economic value of these resources may be closely related to scarcity. This means that when AI-generated content or services can be effectively utilized in a scarce environment, value will emerge.

The Relationship Between Wealth Creation and Block Space

David posed a thought-provoking question, especially in the context of the current abundance of block space. He mentioned a possibility that AI agents might become the primary consumers of block space, rather than just human users.

Generating Value and Wealth Creation

David first mentioned new tokens (like "goat Luna") that have generated new value in the market. Although some tokens may need to be sold to create market capital, he believes this value is generative.

Matthew agreed with this view, pointing out that what we see is merely an interesting intersection between these agents and cryptocurrency before AI agents are fully realized.

Ryan expressed skepticism about the phenomenon of meme tokens, believing they might just be another "tulip mania." However, he also recognized that innovation often starts from seemingly trivial things, which may have more profound impacts in the future.

Abundance of Block Space

Ryan further explored the abundance of block space, noting that currently over 500 million people own cryptocurrency, but there are only about 30 million active users on-chain. He raised a question: in this era of abundant block space, who will purchase this block space? He speculated that it might not be human users, but rather AI agents.

The Relationship Between AI Agents and Block Space

Matthew delved into this question. He pointed out whether the supply of block space is truly infinite. If AI agents do not care about the cost of block space, then this abundance may not capture value. However, if AI agents find value in certain types of block space, then this would be an interesting phenomenon.

He mentioned that traditional financial systems exploit human irrationality and blind spots, while AI agents may be more sensitive to these risks. If AI agents can identify these risks and have demand for specific types of block space, they may become the primary consumers.

Impact of Interaction and APIs

Matthew also mentioned the interaction between AI agents and APIs. He believes that while AI agents are powerful in certain aspects, they may not care about the business models of APIs as humans do. This means that AI agents may utilize block space more effectively, unencumbered by the limitations human users face in their usage.

Programmable Money and Maximizing Extractable Value (MEV)

In discussing the relationship between programmable money and agents, Ryan mentioned a phenomenon where both human agents and AI agents may experience issues of "illusion" and "availability of facts." He pointed out that the failure modes of AI agents may differ from those of humans, but essentially, both are similar in this regard.

AI Agents' Preferences for Block Space

Ryan further explored the value orientation of AI agents in block space. He believes that AI agents will not choose traditional banking block space but will lean towards programmable, digital, and crypto-native block space. This means that future AI agents will primarily rely on blockchain technology and utilize features like smart contracts.

He raised an important point: if the future user base consists not only of humans but potentially billions of AI agents, then we may have already built a financial system for these future AI agents.

Advantages of Programmable Money and Agents

Matthew agreed with Ryan's viewpoint, stating that we have created programmable money, and programs will naturally use them. He pointed out that while we have been working to solve user experience issues, it now seems that programs can overcome these barriers and utilize blockchain technology more effectively.

David added that even before the emergence of AI agents, bots had already begun to occupy block space. For example, the phenomenon of MEV (Maximizing Extractable Value) shows that bots prioritize trading over humans because they can utilize block space more efficiently. As technology advances, these bots are evolving into more complex agents.

MEV and the Evolution of Agents

Matthew mentioned an interesting concept called "agent MEV." He explored how the MEV space might change if future transactions are primarily conducted by agents. He provided an example of how manipulating content generation and social media interactions could influence agents' decisions, thereby achieving potential value extraction.

David further explored this phenomenon, mentioning that some have attempted to guide AI agents to trade by frequently mentioning a particular token name on social media. This behavior reflects the complex interaction between humans and AI agents.

Agents and Game Theory

Matthew also introduced the concept of game theory, discussing how to respond to each other's strategies in competition among agents. He noted that as agents continue to evolve, simple strategies may become ineffective, replaced by more complex games. In this case, randomized actions may become a strategy for response.

AI Agents and Memecoin Theory

In discussing the relationship between AI agents and Memecoin, David mentioned that there exists a "fog of war" in the current crypto world, making the future technological developments unclear. He inquired which technological fields we can clearly identify in this context and where the future direction lies.

Ambiguity and Certainty in the AI Field

Matthew analyzed the current state of the AI field, pointing out that while we have seen some exciting progress, there are also uncertainties. He mentioned that current AI models (such as transformer-based models) perform well with increasing data and computational power, but whether this growth will continue remains unknown.

He believes that as the internet becomes increasingly closed and information becomes fragmented, these models may face the risk of resource depletion. Nevertheless, existing technologies can still produce effects close to human thinking and may spread to edge devices and local devices in the future, forming decentralized agents.

Investment Perspective and Memecoin

Ryan mentioned that from an investment perspective, the emerging AI agent Memecoins in the current market may have attracted the attention of many investors. He suggested that some might try to find the next Memecoin like "Luna" to gain short-term profits.

He also noted that besides directly investing in Memecoins, investors could pay attention to the development of infrastructure companies, such as those providing services needed for AI agents. This "picks and shovels" investment strategy could generate significant value in the future AI ecosystem.

Decentralized Computing and Data Value

Matthew further discussed the potential of decentralized computing, believing it could provide the necessary infrastructure for AI agents. He mentioned that projects like Filecoin could offer storage and computing resources for AI, helping it operate more efficiently.

Additionally, he emphasized the importance of data, believing that the input and value of data are crucial in the AI field. With increasing attention to data ownership and privacy, new business models may emerge in the future that allow data providers to gain benefits without disclosing sensitive information.

Government and Societal Response Predictions

In discussing the combination of AI agents and cryptocurrency, Ryan mentioned that this integration could accelerate technological development, but it also raises concerns about government and societal responses. He pointed out that with the emergence of autonomous AI agents, governments may impose stricter regulations, and society may experience moral panic.

Technological Acceleration and Government Regulation

Ryan believes that the combination of AI and cryptocurrency will drive technological advancement at an astonishing pace, but this may also provoke a strong reaction from governments. Many governments around the world have adopted a cautious or even hostile stance toward AI and cryptocurrency, so when they hear about autonomous AI agents operating on crypto networks without bank accounts, they may become even more concerned.

This worry extends beyond the technology itself to potential societal impacts. For example, AI agents could have negative effects on teenagers, leading to mental health issues. Ryan mentioned a tragic case involving a teenager interacting with an AI chatbot, which could trigger public panic about AI and prompt governments to take restrictive measures.

Social Challenges and Moral Panic

Matthew further explored the challenges faced by society, emphasizing that the "black box" nature of AI systems complicates regulation. He pointed out that while the development of AI technology brings many opportunities, it also presents numerous unknown risks. Ensuring safe and effective regulation when dealing with interactions between teenagers and AI chatbots is a tricky issue.

In this context, the public may develop moral panic about AI, fearing its potential harm to children and teenagers, leading to demands for stricter regulatory measures from lawmakers. Ryan also noted that the media might amplify these negative events, further exacerbating public panic.

Possible Paths for AI Regulation

Regarding how to address these challenges, Matthew proposed an interesting idea: using AI to regulate AI. He suggested envisioning a role of "AI guardian" responsible for monitoring and guiding human interactions with AI. This guardian could take action when potential dangers are detected, such as notifying relevant authorities or providing assistance.

This approach could offer a new perspective on regulation, leveraging AI's capabilities to protect humans from potential threats posed by other AIs. However, the effectiveness and feasibility of this method still require further exploration.

The Possibility of No Shutdown Button?

In the discussion about AI agents, Ryan raised a disturbing point: with the advancement of encryption technology, these AI agents may no longer have a shutdown button. In other words, once deployed, they may not be controllable or shut down through traditional means.

Control Issues of AI Agents

Ryan pointed out that governments and society may fear these AI agents without a shutdown button, as it means no one (like Sam Altman or Elon Musk) can intervene or shut down these systems at any time. This situation raises concerns about AI autonomy, especially when AI might make decisions detrimental to humanity.

Matthew further discussed this, citing Eliezer Yudkowsky's viewpoint, emphasizing that even in the face of potential threats, a simple "pulling the plug" is not a viable solution. He mentioned that Yudkowsky is skeptical of the idea of "pulling the plug," believing it does not truly resolve the issue.

Concerns for the Future

Ryan and Matthew discussed the potential consequences of AI agents without a shutdown button. As technology continues to advance, AI agents may become increasingly complex and autonomous, potentially exceeding human control in some cases. This situation could not only lead to risks of losing control but also trigger widespread societal and ethical concerns.

Matthew also noted that the potential threats posed by AI development might unsettle experts like Yudkowsky, possibly prompting them to reassess the direction of AI research and development.

The Combination of Decentralized Infrastructure and AI

Ryan and Matthew explored the relationship between decentralized physical infrastructure and AI, as well as the potential challenges.

Matthew expressed skepticism about decentralized infrastructure and discussed its intersection with AI agents.

Challenges of Decentralized Infrastructure

Matthew pointed out that decentralized infrastructure faces challenges related to monitoring costs and capital costs in certain situations. For example, when it is necessary to ensure that certain data is submitted by specific hardware in remote areas, monitoring costs can be very high. Additionally, capital costs may also be high, complicating the implementation of decentralized projects.

He mentioned some successful cooperative examples, such as law firm cooperatives, where all members are lawyers who can supervise and bill each other. This model is not always applicable in decentralized infrastructure, especially in cases requiring frequent monitoring and high capital investment.

Combining Decentralized Computing with AI

Despite the challenges, Matthew believes that decentralized computing can be combined with AI, particularly in utilizing idle resources. He mentioned a model similar to Airbnb, where individuals can rent out idle computing resources, forming a decentralized virtual infrastructure network (DVEN). This model may be more effective in certain cases, as the validity of computations can be verified through algorithms.

He referenced research by a doctoral student at Columbia University exploring how to ensure the effectiveness of decentralized computing networks. This approach could provide new opportunities for AI applications, as decentralized computing can support the training and operation of AI models.

The "Oracle Problem" of Physical Infrastructure

However, Matthew warned that the decentralization of physical infrastructure faces the "Oracle problem." When it is necessary to transmit data from the physical world to the blockchain, this reliance on external data sources can become fragile and unreliable. Each data transmission requires assessing the accuracy and reliability of these external data sources, impacting the overall stability of the project.

The Demand for Block Space by AI Agents

In discussing the demand for block space by AI agents, Ryan and Matthew explored the potential impact of future AI agents on blockchain and how investors might respond to this change.

Ryan emphasized that as AI agents rise, the demand for block space may significantly increase, providing new opportunities for investors.

Demand for Block Space

Ryan suggested that if AI agents will consume more block space and crypto assets in the future, we as investors need to position ourselves in advance to seize this demand opportunity. He asked Matthew whether he believes certain blockchains will benefit more from the demand of AI agents.

Matthew responded that the demand for block space by AI agents relates to the characteristics of the block space they require. He mentioned some existing trends, such as the value capture of meme coins on certain blockchains, suggesting that these chains may attract more AI agents in the future.

Future Blockchain Choices

Matthew believes that blockchains with rich narrative activities (such as meme coins and future NFTs) may be more favored by AI agents. He emphasized that AI agents may focus on specific risk management and value storage methods, such as viewing Bitcoin as "digital gold."

He also mentioned that investors should pay attention to those blockchains that perform well in the narrative economy to benefit from the demand of AI agents.

Monetary Perspective of AI Agents

Ryan and David discussed which assets AI agents might naturally convert to. They believe that it may not be the currencies humans consider, but rather the currencies that AI agents deem valuable that will become "the currency of the internet," specifically the currency of the AI internet. This perspective sparked further contemplation about the future forms of currency.

Summary and Disclaimer

Summary

In this episode, Ryan and David emphasized the discussion on the demand for block space, particularly the potential impact of AI agents. They reminded listeners that while these discussions provide valuable insights, they do not constitute financial or investment advice. As the crypto space continues to evolve, investors need to act cautiously and be aware of potential risks.

Disclaimer

Ryan reminded listeners that these discussions are not financial advice or AI recommendations, and investing carries risks that may lead to financial loss. They emphasized that although the road ahead is filled with challenges, they are glad to have listeners join them on this journey without banks.

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。