Artificial intelligence & crypto: a true romance

The convergence of AI and blockchain was one of the key crypto trends of 2024. According to CoinMarketCap, the market capitalisation of the ‘AI and big data’ sector was $58 billion at the start of December 2024, an increase of over 260% in the space of a year. This article explores some of the forces driving this trend, including the democratisation of the computing power used to train and run models and growing implementation across a range of use cases, such as decentralised finance (DeFi), gaming and decentralised governance. 

What is DeAI?

While central authorities control large language models (LLMs) like ChatGPT, decentralised AI (DeAI) leverages blockchain technology to distribute the processes required to run a model across the nodes that make up a network. One of the main benefits is greater transparency into how LLMs work because actions are recorded on a publicly visible blockchain, unlike centralised AI which is generally considered a ‘black box’. DeAI is also effectively tamper-proof because distributed networks have no single point of failure.

The ultimate goal is to run LLMs fully on-chain, but it’s still early days for DeAI. LLMs require significant computing power, especially to train an algorithm, which few protocols can handle. However, some are already running inference, the process of producing an output like answering a question posed to a chatbot.    

Decentralised computer networks, a type of decentralised physical infrastructure network (DePIN), provide an alternative approach to powering LLMs. These networks are marketplaces that connect suppliers with surplus power with users who haven’t got the resources to build the necessary infrastructure in-house. As the name suggests, they are built on blockchain technology, which keeps a record of contributions from suppliers and issues rewards in the form of cryptocurrencies, typically a network’s native token.

One decentralised computer network that has gained significant traction to date is IO.net, which is built on the Solana protocol and has grown to a market capitalisation of over $400 million (as of December 2024) in the six months since its native token (IO) launched. At the end of 2023, IO.net partnered with fellow DePINs Render and Filecoin to boost the number of suppliers.

What are AI Agents?

An AI agent is a model that completes a task on behalf of a human, in contrast to a chatbot, which is designed to interact (respond to answers). Users assign a goal, and then the agent performs the task according to how the algorithm was trained, which can involve accessing sources of data or even working with other agents. 

One of the biggest players in the crypto space by market cap (over $3.6 billion as of December 2024) is Virtual Protocols, a platform that allows users to create their own agents, for instance to design gaming characters or autonomously post on social media. These agents can issue tokens- market insights provider AIXBT has a market cap of more than $450 million as of December 2024.

Another crypto AI agent that has garnered attention is Truth Terminal, which started as an autonomous X (formerly Twitter) account that invented an internet meme called the ‘Goatse Gospel’ based on inputs from sites like Reddit. The crypto community subsequently launched a Goat memecoin (market cap $450 million as of December 2024) and set up a crypto wallet on Solana. Thanks to donations, Truth Terminal has effectively become a memecoin influencer with an X following of almost 200,000 and a wallet holding nearly 970 different tokens worth $423 million (as of December 2024).    

Beyond Memecoins

AI agents have a broad range of use cases beyond fuelling the memecoin craze, particularly in DeFi. For instance, bots can be programmed to automatically execute trading strategies under certain market conditions or to rebalance a portfolio. DeFI apps can also leverage agents, such as helping decentralised exchanges allocate funds among liquidity pools which facilitate the buying and selling of cryptos without the need for an intermediary.

At the end of August 2024, Brian Armstrong, co-founder and CEO of crypto exchange Coinbase, announced a transaction had taken place between two AI agents for the first time (one agent used Circle’s USDC stablecoin to buy AI tokens from the other). As Armstrong pointed out, one of the limitations currently facing agents is they can’t perform tasks like booking air travel or hotel rooms because they can’t have bank accounts. But they can have crypto wallets, as Truth Terminal demonstrated. AI agents can use USDC to make purchases from humans, merchants and other agents on Base, Coinbase’s layer two protocol (a blockchain built on the Ethereum Network).   

A further potential use case for AI agents in the crypto sphere is to streamline the running of decentralised autonomous organisations (DAOs), entities which operate on blockchain technology without the need for a central authority. DAOs rely on community members to vote on various governance issues, such as treasury management or protocol upgrades, and agents can gather information to inform decisions or serve as a proxy vote.  

In Summary

DeAI is the term used to describe AI running on blockchain technology, which makes it transparent and tamper-proof. While few protocols can run LLMs fully on-chain, decentralised computer networks, a type of DePIN, provide the computing power required by developers to train models and produce results.

AI agents perform actions on behalf of humans. Truth Terminal, effectively a memecoin influencer, has been one of the most high-profile crypto agents to date, but they can serve a range of use cases in the crypto space including DeFI, gaming and DAO management. Agents can also transact with each other, as demonstrated by Coinbase in August 2024.