“AI agents on the ICP can operate autonomously within canister smart contracts, executing core business logic directly on-chain.”
Jennifer Tran
An AI agent is a software system designed to perform tasks or make decisions autonomously based on data inputs, algorithms, and predefined objectives. These agents use artificial intelligence methodologies such as machine learning and natural language processing to interact with their environments, analyse data, and take actions without human intervention.
The complexity of AI agents can vary. Some may be simple systems following predefined instructions, while others are more advanced models capable of learning from experience and adapting their behaviour accordingly. AI agents may perform tasks like pattern recognition, problem-solving, or real-time responses to dynamic conditions. Examples of AI agents include virtual assistants like Siri and Alexa.
A concrete example of an AI agent performing pattern recognition, problem-solving, and real-time responses is SimpliSafe’s Active Guard Outdoor Protection system. This security solution utilises an advanced outdoor camera equipped with AI capabilities to monitor properties. The AI agent analyses live video feeds to detect potential threats by recognising patterns indicative of intrusions. Upon identifying suspicious activity, the system enables live agents to assess the situation in real-time and take appropriate actions, such as communicating with the intruder through a loudspeaker or contacting emergency services.
Source: The Verge
AI agents are increasingly utilised in various contexts, with their capabilities extending beyond performing complex tasks. For example, an AI agent could be designed to possess detailed knowledge about the anime TV series
Naruto
. Such an agent could be trained and programmed with comprehensive information related to the series, enabling it to assist users interested in specific aspects of
Naruto.
Jensen Huang, CEO of NVIDIA, has discussed the transformative potential of AI agents in the workforce. He predicts that by the end of 2025, 30% of companies will have digital employees contributing meaningfully to enterprises. Huang envisions that information technology departments will evolve to manage these AI agents, effectively becoming the HR departments for digital workers.
Source: Fortune.com
The evolution of artificial intelligence agents has been marked by a significant progression from early systems, such as Eliza, to more sophisticated frameworks like AI16Z. Eliza, developed in the mid-1960s by Joseph Weizenbaum at the Massachusetts Institute of Technology (MIT), was one of the first natural language processing programs designed to simulate conversation through a methodology of pattern matching and response substitution.
In contrast, AI16Z is a major advancement in AI applications, functioning as a decentralised autonomous organisation (DAO) that facilitates data-driven financial decision-making. AI16Z aspires to revolutionise the investment, governance, and operational frameworks within cryptocurrency communities, employing AI to improve decision-making processes.
At the heart of AI16Z is an AI agent named Marc AIndreessen, which autonomously analyses investment opportunities and executes trades based on comprehensive data evaluation. The AI16Z token plays a crucial role within the ecosystem, acting as both a governance and utility token.
The agent functions on the Solana blockchain while relying on Amazon Web Services for computational resources (as Solana uses AWS for the computation) and as we know notably, the
ICP stands out as a platform capable of executing AI functionalities on-chain
.
Read our articles about the Ai On chain.
Among the notable AI agents is ELNA, which operates within the protocol and offers a tamper-proof environment. ELNA is built exclusively on the Internet Computer Protocol, which facilitates users’ creation, deployment, and monetisation of AI agents directly on the blockchain. This design promotes transparency, security, and community ownership of AI agents.
The platform enables users to upload data that can be utilised to train distinctive machine learning models, focusing on specific topics and applications. ELNA’s architecture incorporates canister smart contracts for the training and deployment of models, vector database architecture for efficient on-chain data storage, and Motoko components for access control, transactions, and analytics.
Vector Database Architecture:
A vector database stores data as vectors, which are mathematical representations used in machine learning and AI applications. These databases are optimised for similarity searches, allowing for quick retrieval and comparison of high-dimensional data. This is especially useful for tasks like image recognition and natural language processing. Vector databases support efficient querying, making them integral to AI systems that rely on fast, accurate data retrieval.
Motoko Components:
Motoko is a language for writing smart contracts on the Internet Computer Protocol (ICP). It enables the creation of canisters, which are computational units running autonomously on the network. Motoko components include actors (entities with state and behaviour), modules (units of code), and functions (logic for interacting with state and other actors). The language is designed for efficiency and safety within the ICP ecosystem.
Source: Dfinity Forum
This architecture ensures that AI agents operate fully on the blockchain, thereby enhancing decentralisation and security.
A significant feature of ELNA is its ability to integrate these AI agents with various external platforms, such as social media (X, Instagram), websites like Wikipedia, or any platform where integration is necessary. The platform simplifies the process by offering tools for on- and off-chain LLMs, a vector database for efficient data storage, a prompt engine for better LLM interaction, and integration capabilities for connecting AI agents to different online environments. This makes it highly versatile in developing autonomous agents that can automate complex tasks, such as real-time analytics, customer support, or even legal and medical assistance.
Creating an AI Agent on ELNA.ai: A Step-by-Step Guide.
AI agents such as ELNA and projects like ONICAI, which leverage the ICP, represent a major evolution in the development of secure, autonomous systems. By operating on-chain, these AI agents benefit from the decentralised, tamper-proof environment provided by the ICP, which drastically reduces the risks associated with centralised platforms.
Centralised AI systems are prone to hacking, data corruption, and manipulation, but the on-chain architecture of the ICP ensures that these agents remain secure, transparent, and immutable.
ONICAI, for instance, has made significant strides by deploying state-of-the-art Large Language Models as on-chain AI, further cementing the security and decentralisation benefits of the ICP. This achievement enables complex AI tasks to be executed directly on the blockchain, offering a pure form of decentralised AI.
The security advantages of on-chain AI agents are immense. With the data and algorithms stored directly on the blockchain, any attempts to alter or manipulate the agent’s behaviour are immediately detected and prevented.
As AI agents like ELNA and ONICAI continue to evolve, they will unlock new possibilities for autonomous systems across a range of industries. By offering secure, transparent, and scalable solutions, these platforms are setting the stage for a future where AI operates safely and reliably, from automating complex tasks to providing digital services in a decentralised environment.
Leave a Reply