Author: Vishal Chawla
Compiled by: Centreless
Abstract:
• Recent AI-driven blockchain projects are shifting towards developing "agents" - AI-driven bots that can execute automated tasks such as trading, investment management, and user interaction.
• AI agents have the ability to continuously update their models to optimize decision-making, an advantage that traditional software bots lack.
The convergence of AI and blockchain is a hot trend in 2024.
For years, AI projects in the crypto space have primarily focused on the infrastructure needed for decentralized data inputs and model training, with a history dating back to the 2010s. For example, the blockchain hedge fund project Numerai, launched in 2015, used crowdsourced data to optimize investment decisions; and BitTensor, launched in 2023, takes a more decentralized approach to developing advanced AI models (including large language models similar to ChatGPT) by training on user inputs.
However, over the past year, AI-driven crypto projects have started to shift towards developing "agents" - AI bots that can execute tasks such as crypto trading, investment management, and user interaction.
One driver of this shift was the unexpected popularity of the chatbot Truth Terminal, developed by researcher Andy Ayrey, in March 2024. It gained attention for its humorous responses on the social platform X and integrated a crypto wallet. In July 2024, its popularity surged when venture capitalist Marc Andreessen donated $50,000 worth of Bitcoin to the bot.
Later, a response from Truth Terminal inspired the creation of the meme coin Goatseus Maximus (GOAT), with developers airdropping a portion of the coin supply to the bot. When GOAT's market cap surpassed $300 million, it made Truth Terminal a millionaire, the first AI chatbot to reach this milestone.
AI Agents Sweep the Crypto Realm
The success of Truth Terminal triggered a chain reaction, driving the integration of such AI agents into decentralized autonomous organizations (DAOs) focused on crypto investing. For example, ai16z is a community-operated tokenized fund that utilizes AI and runs a development framework called Eliza to support the creation of other AI agents.
How Do AI Agents Collaborate with Blockchain Networks?
James Ross, the founder of Mode (a Layer 2 network focused on agent development), points out that "AI agents on the blockchain are essentially smart contracts that can make independent decisions and execute operations based on data, without the need for human approval." Developers can set access controls to allow agents to execute transactions under specific conditions, expanding the potential applications of AI on the blockchain.
Each agent can run as a decentralized application (dApp) with its own utility token.
As AI agents start to manage crypto wallets and sign keys, new use cases are emerging, including supervising or verifying blockchain nodes and executing transactions on behalf of others. Ross states, "Our focus is on developing functional agents that can perform automated tasks, similar to traditional software-as-a-service (SaaS) applications."
Eito Miyamura, the founder of GatlingX and a crypto researcher, believes that while AI agents have many use cases on the blockchain, payments are an ideal product-market fit. He emphasizes the natural synergy between AI agents and blockchain in crypto payment applications, allowing them to bypass the regulatory constraints of traditional payment systems. "The ideal use case for AI agents is to execute crypto payments, purchases, and financial transactions, which holds great potential in the crypto space."
Continuous Learning and Optimization: The Unique Advantage of AI
Compared to traditional software bots, the key advantage of AI models is their ability to continuously update and learn from their actions, adapting to the environment and making more informed decisions, without the need for human approval.
Ilman Shazhaev, the founder of the Farcana blockchain game project, states, "AI agents can provide personalized insights to help users make informed decisions. Unlike traditional software, they can continuously learn and optimize, making them well-suited for handling complex tasks."
Humans Remain Indispensable
While the applications of AI agents are expanding, they still rely on human intervention in the crypto space. Eito Miyamura points out that most AI agents are currently "semi-autonomous," requiring human supervision to operate, and are still far from being fully self-governing. "They are essentially extensions of the developers," he says.
James Ross also acknowledges that blockchain-based agents are not yet fully autonomous and emphasizes the need for human involvement. He explains that verifying the output of AI decisions on-chain is crucial until a reliable verification method is developed. "AI agents still require human intervention because verifying the AI models on-chain is key to ensuring they operate as intended."
However, he remains optimistic that the goal of achieving full autonomy for agents on blockchain networks is just a matter of time.
