Fetch.ai is an open source software project that aims to build infrastructure for developing modern, decentralized, and peer-to-peer (P2P) applications. Leveraging artificial intelligence and automation, Fetch.ai provides a variety of tools and frameworks to create and connect intelligent agents to perform complex tasks in the digital economy. An intelligent agent is an autonomous software code that can act on behalf of a human, organization, or machine. Fetch.ai's network is a cross-chain protocol based on the Cosmos-SDK that enables advanced cryptography and machine learning logic on the chain. Fetch.ai also has its own cryptocurrency called FET, which currently has a circulation of 746 million and a maximum supply of 1.153 billion.
As a technology company that deeply combines blockchain and artificial intelligence technology, Fetch.AI aims to build a decentralized smart economy by combining artificial intelligence, blockchain and IoT technologies to achieve distributed goals. The company's goal is to provide businesses and consumers with a new way to interact economically and achieve more efficient, safer and smarter transactions.
Thanks to the high intelligence and open architecture of AI+blockchain, Fetch.AI has a wide range of application scenarios, including logistics, supply chain, finance, energy, medical care and other fields. Fetch.AI's technical architecture mainly consists of two parts: Fetch.AI main chain and Fetch.AI smart agent. The Fetch.AI main chain is a distributed ledger based on blockchain technology, which is used to record transactions and smart contracts and ensure the security and reliability of transactions. Fetch.AI smart agent is a smart contract with artificial intelligence capabilities that can autonomously perform tasks, coordinate resources and interact with other smart agents, thereby realizing automated, intelligent and decentralized economic interactions.
This article will not go into too much detail about the main chain. We will focus on the Autonomous Agent Architecture (AEA) and Colearn mechanisms to demonstrate how AI participates in the operation and data application of the blockchain system.
Let network nodes manage themselves: Autonomous Economic Agent Architecture (AEA)
On the Fetch.ai network, individuals or companies with data are represented by their agents, who connect with agents of individuals or companies seeking data. Agents operate on the Open Economic Framework (OEF). This acts as a search and discovery mechanism where agents representing data sources can advertise the data they have access to. Similarly, individuals or companies seeking data can use OEF to search for agents that have access to the relevant data.
Fetch.AI's AEA architecture is a distributed intelligent agent architecture used to build an autonomous and collaborative intelligent agent network. AEA stands for Autonomous Economic Agent. Its core idea is to combine artificial intelligence and blockchain technology to build a decentralized intelligent economy and realize intelligent, autonomous and decentralized economic interaction.
The core components of the AEA architecture mainly include the following four modules:
AEA Agent: AEA Agent is an autonomous, programmable intelligent agent with the ability to make decisions, collaborate and learn independently. It is the core component of AEA and represents an independent entity with the ability to make decisions and act independently. Each AEA agent has its own wallet address, identity and smart contract, and can interact and cooperate with other agents.
AEA Communication (Connection): AEA communication is a peer-to-peer communication protocol based on blockchain technology, which is used to realize information transmission and interaction between agents. AEA communication can ensure the security and reliability of interaction. Fetch.AI's AEA supports multiple connection methods, including WebSocket and HTTP connections.
AEA Skill: AEA Skill is a pluggable module that extends the functionality and capabilities of AEA agents. Each skill consists of a smart contract and a Python package that implements specific functionality of the agent, such as natural language processing, machine learning, decision making, etc. Skills can contain multiple protocols and models so that agents can understand and respond to requests from other agents.
AEA Protocol: AEA protocol is a collaboration mechanism for achieving collaboration and interaction between agents. The AEA protocol defines the message format, protocol flow, and interaction rules between agents, thereby achieving collaborative work between agents. The protocol is the rules and guidelines for communication between agents. The protocol defines how agents should exchange information, respond to requests, and handle errors. Fetch.AI's AEA supports multiple protocols, including Fetch.AI's own Agent Communication Language (ACL) and HTTP protocol.
Imagine a company is seeking data to train a predictive model. When the company’s agent connects to the agent representing the data source, it will ask it for information about the terms of trade. The agent, working on behalf of the data provider, will then provide the terms under which it is willing to sell the data. The agent selling access to the data may seek the highest possible price, while the agent buying access to the data wants to pay the lowest possible price. However, the agent selling the data knows that it will miss out on the deal if it charges too high a price. This is because the agent seeking the data will not accept those terms and will instead try to buy the data from another source on the network. If the purchasing agent does find the terms acceptable, it will then pay the agreed-upon price to the selling agent via a transaction on the Fetch.ai ledger. Upon receiving payment, the agent selling the data will send the encrypted data across the Fetch.ai network.
Aside from the initial setup, the entire process is fully automated and performed by the Fetch.ai agent. This means company employees are able to work without interruption while the predictive model accumulates relevant, anonymous data. By fetching the data, the company purchasing the information is able to train its model more efficiently, which can then be used to make more accurate predictions. Such predictions can be used in any industry.
The core of making nodes intelligent: AEA skill module and group learning (Colearn) mechanism
Among the above four modules, the most important one is the AEA skill module, which is the key module for making nodes intelligent. AEA skills are pluggable modules used to implement group autonomous learning functions of agents. Each learning skill includes a smart contract and a Python package to implement different types of learning tasks, such as reinforcement learning, supervised learning, unsupervised learning, etc. When an agent needs to learn, it can choose a learning skill that suits it and save the learning results in its own state. Agents can autonomously adjust their behaviors and strategies based on the learning results, thereby achieving smarter, more efficient and more sustainable economic interactions.
Fetch.AI’s collective learning principle includes the following steps:
Data sharing: Different agents collect their own data and upload it to a shared database in the blockchain network. This data can be sensor data, text data, image data, etc. All agents participating in collective learning can access the data in the shared database and use this data for training.
Model training: The agent uses data from a shared database to train a model. The model can be a machine learning model, a deep learning model, or another type of algorithm. The agent can be trained with different models to learn different tasks or problems.
Model selection: After model training is completed, the agent uploads its model to the blockchain network. All agents participating in collective learning can access these models and choose the model that suits them according to their needs. The selection process can be based on factors such as agent performance, task requirements, resource constraints, etc.
Model integration: After selecting a model, the agent can integrate it with its own skills to better complete its tasks. Skills can be modules that handle specific types of tasks, such as cryptocurrency trading, logistics management, etc. Agents can use multiple skills and models for task processing.
Reward mechanism: In the process of collective learning, agents can get rewards by contributing their own data and models. Rewards can be distributed based on factors such as agent performance, contribution, resource utilization efficiency, etc. The reward mechanism can encourage agents to actively participate in collective learning and improve the performance of the entire system.
Suppose there are two agents A and B, who need to cooperate to complete a task, such as transporting goods. Agent A is responsible for providing goods, and agent B is responsible for providing transportation services. In the initial interaction, both agent A and agent B can adopt random behavior strategies to complete the task, such as randomly selecting a transportation route or mode of transportation.
As the interaction proceeds, Agent A and Agent B can learn the interaction history data through learning skills and autonomously adjust the behavior strategy according to the learning results. For example, Agent A can learn information such as the supply and transportation cost of goods through learning skills, so as to autonomously select the optimal cooperation strategy according to the current demand for goods and market prices. Agent B can also learn information such as the efficiency and cost of transportation routes and transportation methods through learning skills, so as to autonomously select the optimal transportation strategy according to the current traffic conditions and energy prices.
As the interaction continues and the learning results are continuously updated, Agent A and Agent B can gradually optimize their behavior strategies to achieve more efficient, smarter and more sustainable economic interactions. This autonomous learning process can be continuously iterated and optimized to achieve better economic benefits and social value.
It should be noted that the autonomous learning function requires the agent to have sufficient computing power and data resources to achieve good learning results. Therefore, in practical applications, it is necessary to select appropriate learning skills and resource configurations based on the actual situation and needs of the agent to achieve the best learning effect.
Fetch.ai's core autonomous economic agent (AEA) achieves the goals of intelligence, autonomy and decentralization in economic interaction. Its advantages lie in the deep integration of artificial intelligence and blockchain technology, as well as the design of autonomous economic agents. These AEA agents can autonomously learn, make decisions and interact freely in a decentralized environment, improving the efficiency and intelligence of economic interactions. In addition, Fetch.AI's group learning (Colearn) mechanism encourages agents to actively participate and improve the performance of the entire system by sharing data and models.
However, Fetch.AI also faces some challenges. First, its autonomous learning function has high requirements for computing power and data resources, which may limit its application in resource-constrained environments. Second, Fetch.AI's technical architecture and functions are relatively complex, requiring higher technical thresholds and learning costs, which may affect its widespread application.
Summary
Looking ahead, Fetch.AI still has a bright future. As technology continues to develop, it may introduce more AI and blockchain technologies to improve performance and efficiency and meet more application scenarios and needs. At the same time, as privacy protection and data security are increasingly valued, Fetch.AI's decentralization and security features may receive more attention and application. Despite some challenges, Fetch.AI's innovation and potential in the fields of AI and blockchain are still worthy of our attention and exploration.
references:
[1] Fetch.AI Developer Documentation
[2] Melanie Mitchell: AI 3.0
[3] Alexey Potapov: Basic Atomese Features required
Disclaimer: This article is for research reference only and does not constitute any investment advice or recommendation. The project mechanism introduced in this article only represents the author's personal views and has no interest in the author or this platform. Blockchain and digital currency investment has extremely high market risks, policy risks, technical risks and other uncertainties. The price of tokens in the secondary market fluctuates violently. Investors should make decisions prudently and bear investment risks independently. The author of this article or this platform is not responsible for any losses caused by investors using the information provided in this article.