Dialogue with Pantera Research Partners: Artificial Intelligence Will Reshape the Crypto Economy, A New Game of Asset Scarcity and Technological Abundance

Original Title: The Rise of AI Memecoins & What It Means For Crypto

Source: Bankless

Translation: 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

Broadcast Date: October 30, 2024

Background Information

The collision between 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 Artificial Intelligence Narratives

• Matthew mentioned that the narrative around cryptocurrency and artificial intelligence 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 decentralized commitment devices (i.e., blockchain) by AI agents. 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 pointed out 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 and can perform tasks according to a certain degree of will without needing explicit programming.

• Ryan further explored the definition of economic agents, asking Matthew whether he, his company (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 gave 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 noted 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, suggesting that agents might need some form of intelligence and goal-achieving capability, while Matthew emphasized that agents are more based on relationships between people rather than being mere 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 artificial intelligence 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 story.

The Rise of GOAT Meme Coin

• Matthew outlined the background of the GOAT meme coin, mentioning that a few months ago, an account began interacting on social media and gradually became interested 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 tweeting to drive its price.

The Impact of AI Agents

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

The Mechanism of AI Agents

• 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 culturally relevant content related to memecoins and interacting with users. The AI posts 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 the 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 using its own tokens. These AI agents still have limitations in interacting with the world, but the emergence of these derivative projects seems to herald more innovation to come.

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 exist on the blockchain. Under smart contracts, there is no difference between AI and humans. Most importantly, AI can accumulate and control its 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 visionary, 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 transferring funds and making tip payments, involving hundreds of millions of dollars in transactions. He pointed out that the self-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 can make tip payments.

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 instability of AI agent behavior, caution is needed when interacting with the blockchain.

The Result? Is This Our Life?

• Ryan expressed amazement 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, noting 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 delineation of legal responsibility is still unclear.

• David noted that as we create autonomous blockchains and smart contracts, the emergence of AI agents could lead to a "Cambrian explosion" phenomenon. He mentioned 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 constrained, while people may hope that AI agents can autonomously generate more exciting outputs. This contradiction between autonomy and constraint fills people with imagination and anticipation for the future of AI agents.

Exciting Use Cases

• Ryan discussed the potential various applications of 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 market and gain wealth by supporting these projects. He envisioned a scenario where users could request graphic generation on social media through AI agents and pay in cryptocurrency, providing powerful functionality for AI agents.

Matthew's Perspective

• Matthew further explored the potential use cases of 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 virtual services sector. According to a McKinsey report, it is estimated that about 20% of the global GDP (approximately $70 trillion) could be accomplished virtually, providing a massive market for AI agents' applications.

Transformation of the Service Economy

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

• Matthew noted that narratives play an important role in the economy and may influence 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 specialization and the construction and destruction of narratives.

Sam Altman's Quote and Its Importance

• 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, while the latter emphasizes scarcity and limitation.

Comparison of Economic Models

• Matthew further analyzed the profound implications of this statement. He pointed out that while AI's creative capacity brings 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 valuable due to 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 may lead investors to overlook its potential value. He emphasized that what is truly valuable is often scarce resources, not widely available abundance. Therefore, understanding the relationship between scarcity and abundance is crucial when considering investments.

The Intersection of Scarcity and Abundance

• Matthew believes that the intersection of scarcity and abundance may provide us with new perspectives on value. For instance, on 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 Wealth Creation Process and the Relationship with 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 are generating new value in the market. While some tokens may need to be sold to create market capital, he believes this value is generative.

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

• 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 far-reaching impacts in the future.

The 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 posed 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, it would be an interesting phenomenon.

• He mentioned that traditional financial systems operate by exploiting 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.

Interaction and the Impact of 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, without being constrained by human users 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 face 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' Preference 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.

The 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 instance, the phenomenon of MEV (Maximally 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 introduced an interesting concept of "agent MEV." He explored how the MEV space would change if future transactions were primarily conducted by agents. He provided an example of how to influence agents' decisions through manipulating content generation and social media interactions to achieve potential value extraction.

• David further explored this phenomenon, mentioning that some people attempt to guide AI agents to trade by frequently mentioning a certain 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 mentioned that as agents continue to evolve, simple strategies may become ineffective, replaced by more complex games. In this case, randomization of actions may become a strategy for coping.

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 future technological developments unclear. He asked what technological areas we can clarify 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 see some exciting progress, there is also uncertainty. He mentioned that current AI models (like 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 results 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 attract 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 that provide services needed by AI agents. This "picks and shovels" investment strategy may create 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 provide storage and computing resources for AI, helping it operate more efficiently.

• Additionally, he emphasized the importance of data, stating that in the AI field, the input and value of data are crucial. As concerns about data ownership and privacy increase, new business models may emerge that allow data providers to earn revenue 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 also raise 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 progress at an astonishing pace, but this may also provoke strong reactions from governments. Many governments have taken a cautious or even hostile stance towards 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 concern extends beyond the technology itself to potential societal impacts. For example, AI agents may have negative effects on teenagers, leading to mental health issues. Ryan mentioned a tragic case involving teenagers interacting with AI chatbots, which could trigger public panic about AI and prompt governments to take restrictive measures.

Societal Challenges and Moral Panic

• Matthew further explored the challenges society faces, emphasizing that the "black box" nature of AI systems complicates regulation. He pointed out that while the development of AI technology presents many opportunities, it also poses many unknown risks. Ensuring safe and effective regulation in handling interactions between teenagers and AI chatbots is a tricky issue.

• In this context, the public may experience moral panic about AI, fearing its potential harm to children and teenagers, leading to demands for stricter regulatory measures from legislators. Ryan also mentioned 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 of using AI to regulate AI. He mentioned 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 provide 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 approach still require further exploration.

The Possibility of No Off Switch?

• In discussing AI agents, Ryan raised a disturbing point: with the advancement of encryption technology, these AI agents may no longer have an off switch. 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 AI agents without an off switch, as this means no one (like Sam Altman or Elon Musk) can intervene or shut down these systems at any time. This situation raises concerns about the autonomy of AI, especially when AI may make decisions that are detrimental to humans.

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

Concerns for the Future

• Ryan and Matthew discussed the potential consequences of AI agents without an off switch. 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 mentioned 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 and can supervise and bill each other. This model is not always applicable in decentralized infrastructure, especially in cases requiring high-frequency monitoring and significant capital investment.

The Combination of Decentralized Computing and AI

• Despite the challenges, Matthew believes that decentralized computing can be combined with AI, especially 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 transfer data from the physical world to the blockchain, this reliance on external data sources can become fragile and unreliable. Each data transfer requires assessing the accuracy and reliability of these external data sources, impacting the stability of the entire project.

The Demand of AI Agents for Block Space

• In discussing the demand of AI agents for block space, Ryan and Matthew explored the potential impact of future AI agents on blockchain and how investors can 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, as investors, we 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 replied that the demand of AI agents for block space relates to the characteristics of the block space they require. He mentioned some current 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 activity (like 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.

The Currency Perspective of AI Agents

• Ryan and David discussed what assets AI agents might naturally convert into. They believe that it may not be the currencies humans consider, but rather the currencies that AI agents deem fit that will become "the currency of the internet," i.e., the currency of the AI internet. This perspective sparked further contemplation on 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 advice; investing carries risks and 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.