Oracle networks are infrastructure that connect blockchains to external systems, thereby enabling smart contracts to execute based on external inputs and outputs. Oracle networks are extremely versatile in the smart contract use cases they can be applied to since they can reliably relay information derived from decentralized consensus based on any set of inputs or outputs.

With the emergence of generative AI, large language models (LLMs), and AI agents that can perform tasks on behalf of their users, the role of oracle networks can be expanded through their application to AI-powered technologies.

As a truth machine providing security and cryptographic truth to a wide variety of smart contract use cases, Chainlink continues to evolve, presenting new opportunities to enhance the reliability of AI responses and unlock novel smart contract use cases at the intersection of blockchain and AI. 

This blog explores two key ways that oracle networks can interact with AI models.

Interfacing With Individual AI Models

Chainlink’s initial use case was aggregating many data sources about market data and feeding that verified, aggregated, high-quality pricing information into smart contracts in the DeFi ecosystem. This security-first, principled design is what created the reliability that enabled DeFi to grow from a sub-$100 million niche ecosystem to a $200+ billion onchain economy.

As the universal blockchain connectivity standard, Chainlink can interface with individual AI models as a single source of inputs into smart contracts, thereby connecting smart contracts to AI. For this to work well, the particular AI model being used must be considered a highly reliable and accurate source of information. 

If an AI model is deemed reliable enough to control value, Chainlink can enable it to interact with a smart contract. In addition, using Chainlink CCIP can enable the AI model to interact with multiple smart contracts on multiple chains at the same time on behalf of the user.

If the individual AI model can give useful responses, but there is some risk that it may hallucinate and give the wrong response or take the incorrect action, oracle networks can be used to introduce an additional level of reliability and redundancy. This is where the second potential use case covered in this post comes into play.

Aggregating the Results of Multiple AI Models

Chainlink oracle networks could be used to aggregate the results of multiple AI models. This would involve connecting them to multiple individual AI models and aggregating their responses within a decentralized oracle network (DON) to get them to come to consensus on a certain threshold that creates a greater degree of reliability in the response.

In such an architecture, the end user is not getting an output from a single AI model—they are getting output from the universe of AI models.

Oracle networks can aggregate responses from multiple AI models to increase reliability.

Getting the answer from a wider variety of AI models might be preferable in some cases because it can create more reliable and accurate responses or reduce the risk of hallucination, making oracle-enabled AI models a superior way for smart contracts to interact with certain information.

Depending on the output of a generative task, the challenge of aggregating inputs from multiple AI models can range from simple to unsolved, and Chainlink can develop solutions addressing common use cases and continue pushing the state-of-the-art to address complex ones.

Applying the Truth Machine to AI Use Cases

Using oracle networks to connect AI models to smart contracts has similarities to what Chainlink has already been doing within the blockchain ecosystem. Chainlink Price Feeds leverage a multi-layered decentralized aggregation system that mitigates single points of failure and helps ensure each oracle report reflects the true market-wide price of assets.

This has enabled Chainlink to enable $12+ trillion in transaction value for other types of data, such as market data. If one views AI models as merely a source of inputs and data, Chainlink oracle networks are directly applicable to them as another type of input that smart contracts can use to power end-to-end automated economic processes.

Connecting AI models to smart contracts can unlock a wide range of new use cases for security, supply chain management, authenticity verification, data storage, and more. If you’d like to explore more use cases on the convergence of AI and blockchain, read Use Cases of AI in Blockchain.