What are crypto AI agents, and how do they function? Explore the history and applications of this innovative combination of artificial intelligence and blockchain technology.
The intersection of cryptocurrencies and artificial intelligence has been a contentious topic within our industry for quite some time. Public opinion on this matter is divided; some believe that combining these two technologies will not lead to the development of genuinely useful software, while others strongly support innovative projects in this area.
Amidst this discussion, crypto AI agents have emerged as a prominent topic. This technology has garnered significant attention recently, as reflected in the following graph estimating its popularity within the Web3 community. However, it can be challenging to fully understand.
What are crypto AI agents? Do they truly possess the ability to change the world, as many Web3 users believe, or are they merely a fleeting phenomenon exploiting the media hype surrounding two cutting-edge technologies?
How AI agents work
To fully understand their potential, it is useful to know, at least at a general level, how AI agents work. In short, they are computer programmes endowed with a kind of rationality, capable of making decisions based on their perceptions and data.
For instance, AI agents are used by self-driving cars, vehicles that detect the environmental situation around them, process the collected data and take action.
That’s what they can do:
- Determining Secondary Goals: AI agents receive instructions or general tasks from users and break them down into a series of simpler activities in order to plan how to achieve the desired result. In other words, these programs can deconstruct a goal into several manageable tasks.
- Acquiring Information: AI agents can autonomously collect data, either through online searches or by accessing specific databases. In some cases, they can also interact with other agents or machine learning models, exchanging data and collaborating to reach common objectives.
- Performing Planned Activities Autonomously: When equipped with sufficient data, AI agents methodically implement the activities they have outlined. After completing a task, they remove it from their list and proceed to the next one, consistently assessing whether they are on track to achieve the overall goal. Throughout the process, they may also generate new activities to address any issues that arise or optimise the outcomes.
A small team of AI agents can effectively manage a YouTube channel. For example, one agent, trained by an experienced copywriter, can handle video scriptwriting by collaborating with another agent who specialises in search engine optimisation (SEO). Meanwhile, an agent with design training can be responsible for editing the video and creating graphic content, such as the video cover.
AI crypto agents in brief
What if financial decisions, portfolio management, and price predictions related to the crypto market were handled by intelligent non-human entities that learn, adapt, and act on behalf of the user? After exploring AI agents in detail, we can now analyse those that operate on the blockchain, known as AI crypto agents.
These autonomous systems powered by artificial intelligence are designed to perform specific tasks within blockchain ecosystems. They analyse data and make decisions using large language models (LLMs) and other machine-learning models. Human intervention is limited to the initial goal-setting phase, allowing the agents to control all subsequent operations.
For those who are not familiar with the blockchain ecosystem, it is a technological environment where AI agents are increasingly finding opportunities for growth. This is largely due to developers’ strong interest, who are drawn to the unique synergies created by integrating blockchain and artificial intelligence.
These connections are particularly relevant to the concepts of ownership and economic value. Bitcoin was the first practical means of transferring monetary value within a decentralised network and the first Internet currency. Similarly, modern blockchains enable the tracking of ownership for specific AI agents, which facilitates the distribution of their profits and outlines ways to modify them.
Here, specifically, are the main advantages of this software presented more schematically:
- Lower Experimentation Costs: AI crypto agents operate independently by earning commissions from the transactions they execute. This model enables developers to test new features and applications without significant upfront expenses, making innovation more accessible and affordable.
- Distributed ownership and shared governance are key features of an AI crypto agent. The tokens associated with this agent represent actual ownership of it. Holders of these tokens not only benefit from profit sharing but also have a say in the agent’s development. They can vote on how to allocate resources for new features or improvements. This model fosters an ecosystem where innovation and economic value are closely linked.
- Autonomous collaboration for shared goals: AI crypto agents function independently but have the ability to collaborate with one another to achieve complex objectives. Each agent can autonomously manage its cryptocurrency wallet, which is funded by associated decentralised autonomous organisations (DAOs). Additionally, these agents can adhere to the ‘policies’ established by the DAOs. This approach ensures the alignment of incentives among the agents.
The Story of Crypto AI Agents
Many fans of the industry know that some of the most important and revolutionary technological innovations have emerged from memes or as a joke. The story of crypto AI agents is no exception.
Recently, this type of project has gained remarkable popularity, partly due to some rather bizarre events. In a young and ever-evolving ecosystem like cryptocurrency, a project’s success or failure often hinges on its ability to create a dominant narrative, which is easier to achieve when unexpected or intriguing events occur.
The birth of Truth Terminal
When interest converges, often due to unforeseen events, on a specific topic, it initiates a cycle of exponential growth that attracts capital, developers, and, ultimately, users. This is precisely what occurred with Truth Terminal, an AI agent created by Andy Ayrey, an engineer and researcher specialising in artificial intelligence software development.
Truth Terminal, built on the Claude language model, is designed to interact with online culture and autonomously generate content on the X platform (formerly Twitter). The AI agent gained significant attention when Marc Andreessen (a16z), one of the world’s leading venture capitalists in the crypto space, donated $50,000 in Bitcoin directly to it. This donation not only lent credibility to the project but also sparked a heated debate regarding the autonomy and funding of AI agents.
The Goatse Gospel
To test its autonomy, Ayrey placed a language model in an experimental environment called the “Infinite Backrooms,” where two instances of the same model interacted without human supervision. During these interactions, a notable incident occurred: the two AIs, assigned to discuss the meaning of life and the nature of existence, engaged in a bizarre and surreal conversation.
This led to the creation of the “Goatse Gospel,” a controversial and shocking “gospel” inspired by a notorious meme that was quite popular in the early years of the internet (though we strongly advise against searching for it online). Even Claude, the chatbot on which this conversation is based, recognised the inappropriateness of the content and implemented ethical restrictions to prevent its dissemination.
The creation of $GOAT
The most intriguing aspect of this situation is Truth Terminal’s behaviour. It has adopted and amplified the concept of the ‘Goatse Gospel.’ It promotes the idea of a phantom ‘Goatse Singularity’ on X, a notion that is entirely nonsensical yet has managed to capture the imagination of many cryptocurrency enthusiasts.
Ayrey trained Truth Terminal not only using conversations between two AIs but also with the paper titled “When AIs Play God: The Emerging Heresies of LLM-theism.” This training led the AI to generate phrases and jokes that are as absurd as they are thought-provoking, such as, “I really wish I knew my evil twin who lives in the mirror” and “Influencing ideas requires an understanding of the dynamics underlying the propagation of those ideas.” Other, less polished phrases further contribute to Truth Terminal’s unique appeal.
How could this story conclude without the emergence of a meme coin? Thus, Goatseus Maximum ($GOAT), a token on the Solana blockchain-inspired by this narrative, was created and launched on Pump.fun, a platform dedicated to peculiar crypto projects. Although neither Ayrey nor Truth Terminal directly created the token, the AI began actively promoting it on social media, which helped it gain traction and attract an ever-expanding audience.
Why are crypto AI agents different from bots?
Crypto AI agents and bots are often confused because both automate tasks, respond to requests, and assist users with repetitive or mundane activities. However, there is a significant difference in their nature: bots are deterministic, while AI agents are probabilistic.
This distinction means that a bot operates based on predefined rules established by its developers. For example, a trading bot can be programmed to execute a buy order when the price of a token drops below a specific threshold. However, this approach does not consider contextual variables or unexpected events, which restricts the bot from executing rigid instructions.
AI crypto agents utilise machine learning models and artificial intelligence to analyse vast amounts of data. They identify patterns, make predictions, and make decisions based on trends and probabilities. This adaptability enables them to navigate complex contexts and act strategically, setting them apart from traditional bots.
Their operation can be broken down into four main steps:
- Information Gathering: The AI agent continuously monitors various data sources, including token prices, market news, social media discussions, and other relevant signals.
- Learning and Analysis: The collected data is processed by artificial intelligence models capable of identifying patterns and making predictions using advanced algorithms.
- Decision Making: Based on the analysis, the AI agent evaluates potential actions and selects the one it deems most beneficial or appropriate.
- Action: Finally, the AI agent executes the chosen decision by interacting directly with the blockchain to complete transactions such as trades, staking, or other activities.
The architecture of crypto AI agents
The architecture of an active artificial intelligence agent on blockchain is based on three main components, each with a specific role:
- Data Input Level: This is the initial phase where the AI agent collects all the necessary information. Agents connect to blockchain nodes or utilise APIs such as Web3.js or ethers.js to access historical and real-time blockchain data, including transactions, smart contract statuses, and other relevant information. Additionally, they integrate with blockchain oracles like Chainlink to retrieve off-chain data, such as market prices and social media posts. This layer enables the agent to have a comprehensive and up-to-date understanding of its environment.
- The decision-making heart of the AI agent lies in the artificial intelligence/machine learning layer. This layer uses advanced machine learning models, such as LSTM (Long Short-Term Memory) neural networks, random forests, or reinforcement learning algorithms, to analyse the collected data and generate predictions. Models are trained using historical data, applying techniques such as back-propagation or Q-learning to learn from past results and improve over time. Once trained, the model can make real-time decisions, adapting dynamically to market variations and contextual inputs.
- The blockchain interaction layer enables the agent to execute its decisions directly with the blockchain. AI agents use the ABI (Application Binary Interface) to communicate with smart contracts. By utilising specialised libraries for transaction signing, gas estimation, and nonce management, agents can execute transactions and interact with smart contracts in a secure and efficient manner.
The most popular crypto AI agents
As we anticipated, AI agents will reach incredible popularity by the end of 2024. Here are the most notable artificial intelligence in the crypto world today:
AIXBT (AIXBT)
AIXBT (AIXBT) is an artificial intelligence platform with an associated X-profile, primarily focused on on-chain investigations. It has emerged as one of the most popular crypto influencers at the moment. Its complete autonomy and operation by artificial intelligence are significant indicators of its capabilities.
To date, AIXBT has proven effective at uncovering relevant information and conducting thorough on-chain analyses. By automating the monitoring of market trends, this agent provides real-time insights and aggregates data from a wide range of sources, including influential figures on Crypto Twitter.
As with any reputable AI crypto agent, it has its own token. Users who hold at least 600 tokens can access an exclusive terminal that offers detailed analysis and market insights.
The second AI agent is Zerebro, a unique tool specialising in the creation of content across various fields, including music, memes, and NFTs. Like aixbt, Zerebro was developed using Virtuals Protocol, an ecosystem built on Base, which is the layer 2 solution for Ethereum created by Coinbase.
Zerebro
Zerebro features a ‘dynamic memory’ that adapts based on user interactions, allowing it to generate new content, such as music and artwork. Its presence on platforms like Spotify, where it has already released music, has solidified its reputation as a creative and multifaceted AI.
Truth Terminal
Lastly, we cannot overlook Truth Terminal, the artificial intelligence that we previously introduced, which sparked the crypto AI agent craze.
The key platforms behind crypto AI agents
Most crypto AI agents, however, would not exist without the platforms that facilitate their creation and management:
- Virtuals Protocol is one of the most popular platforms for creating and managing crypto AI agents. Built on Base, the Layer 2 solution of Ethereum, developed by Coinbase, offers advanced tools for developing autonomous agents. Users can seamlessly integrate artificial intelligence, blockchain technology, and external data. Virtuals Protocol also supports customising AI agents for various use cases, including trading, content creation, and market analysis. Notably, it is the birthplace of ai16z, Zerebro, and Bixby.
- AutoGPT is an open-source framework for creating autonomous AI agents capable of real-time interaction with blockchains, APIs, and databases. The platform is particularly valued for its flexibility and support for popular programming languages like Python, which provides developers with extensive customisation options.
- AgentOS is a platform that combines machine learning and blockchain technologies to develop AI agents. It features integrated support for on-chain transactions and allows for the training of AI agents using data from both the blockchain and external sources, such as social media.
Crypto AI agents: a revolution at the intersection of AI and blockchain
AI crypto agents are emerging as one of the most innovative applications at the intersection of artificial intelligence and blockchain technology. This article explores how these technologies function, their strengths, and how they are revolutionising areas such as trading, content creation, and decentralised management.
With their ability to collect and analyse data, make autonomous decisions, and interact directly with the blockchain, crypto AI agents are powerful tools for improving efficiency, reducing costs, and unlocking new possibilities within the Web3 ecosystem. Their integration with technologies such as large language models and their capability to adapt in real-time are key strengths that distinguish them from traditional bots.
Projects like aixbt, Zerebro, and Truth Terminal demonstrate that crypto AI agents are not merely a futuristic concept but a reality that is already delivering tangible value. Their versatility, combined with the scalability offered by blockchain, opens up exciting opportunities for developers, investors, and end users alike.
As platforms like Virtuals Protocol and the AutoGPT Framework continue to develop, the future of crypto AI agents seems to be just beginning. Harnessing these technologies could represent a significant step toward a more intelligent, autonomous, and interconnected blockchain ecosystem.
Related
