Placeholder: Artificial intelligence is on-chain

By Joel Monegro, Placeholder; Compiled: Pine Snow, Golden Finance

As the production cost of AI models decreases, the number of AI agents will grow exponentially. Agents will soon be online more than humans, creating, consuming, and exchanging more information than we can imagine. But if we get a million-fold increase in digital activity, with 99% of that growth coming from machines, we will struggle to cope with this shift without adopting blockchain infrastructure and business models that need to give agents their full potential while allowing us to identify, control, and audit their actions.

Today, companies like OpenAI bear the huge cost of producing models and then selling access through their proprietary interfaces and APIs, and they are mostly limited to consuming and creating content. But to make the most of AI’s potential, we need a large number of professional agents who can talk and trade with each other. They must be able to roam the internet freely and be able to own and spend money on our behalf (or themselves) to perform tasks. We also need ways to identify, control, and audit their behavior.

The problem is that we can’t force agents to comply with our laws, which hinders effective regulation. They cannot use the traditional financial system that relies on the jurisdiction’s identity model, limiting their ability to transact. They consume vast amounts of online information, imposing all traffic costs on service providers, but do not generate revenue from subscriptions or clicks on ads. To solve these problems, we need a digital-native legal and financial system, combined with new business models, to take full advantage of the opportunities presented by this new technology. **

The solution requires (1) a sovereign digital infrastructure with a new software paradigm that guarantees trusted code execution with an immutable audit trail; (2) an independent digital financial system that treats man and machine equally; (3) A cryptographic identity model combined with decentralized communication and reputation protocols. This can only be achieved using blockchain protocols and smart contract applications. **

Blockchain protocols offer a variety of decentralized digital services that can be accessed through smart contracts and paid for with digital assets. For example, smart contract networks such as Ethereum and Solana can securely and reliably execute open-source software backed by auditable trails of blockchain transactions, while networks such as Filecoin and Arweave offer inexpensive and scalable on-chain data storage services. As it becomes easier and cheaper to build new protocols on existing platforms, the range of services offered through decentralized networks is expanding.

We can use these platforms to train, deploy, and operate agents in a decentralized way, but more importantly, the ability to consume them through smart contracts facilitates interaction between agents. It is much harder for AI agents to consume typical Web2 REST APIs, while it is easier to read smart contracts and pay for their services using tokens, without the need for an account or credit card.

Wallet-enabled agents can use any smart contract service or platform, from infrastructure services to DeFi protocols to social networks, unlocking a range of new capabilities and business models. Agents can pay for their own resources as needed – whether it’s compute or data resources – which is a good idea. It can trade tokens on decentralized exchanges for different services or leverage DeFi protocols to optimize its financial operations by lending opportunistic loans or earning income from its assets. It can vote in DAOs and even charge token fees for its features and trade information with other specialized agents in exchange for money. The result is a large and complex economy with specialized AI agents communicating with each other through decentralized messaging protocols and on-chain transaction information, while bearing the necessary costs. This is not possible in the traditional financial system.

Consider the consequences of this idea: If agents act on-chain — even if they think they are off-chain — we end up with a public, immutable, cryptographically signed record of their activity over time. Blockchain will ensure the secure deployment of AI at scale, allowing us to perform things such as auditing agents’ internet behavior, distinguishing between machine-made and artificial content, and building identity and reputation systems for machines based on their on-chain activity. It will help us and them identify and reward good participants (with tokens or reputation), punish bad participants (e.g., by cutting), and judge which agent performs better than another on a particular task. Agents will then be able to decide who to rely on based on their on-chain history, which they can easily access due to the open-source nature of blockchain data.

To achieve this vision, much of the work is coming together. Thanks to new consensus mechanisms and scaling solutions, blockchain infrastructure is rapidly becoming fast and cheap. Smart contract wallets and “wallet-as-a-service” (WaaS) providers will enable agents to transact; At the same time, emerging account abstraction techniques can allow for interaction between humans and agents, and we can authorize agents to make spending from our wallets. We can use the agent’s public key as an identifier to build the registry and reputation system (including, for example, blocklists or cutback mechanisms). We can even play with DAO-owned agents and try new business models; Perhaps some agents will live on their own Layer 2 networks, owned and managed by a distributed community of operators.

These ideas may seem far-fetched at first, but in hindsight they are very obvious. Of course, smart contracts will mediate business activities between agents, just as legal contracts govern human activities. Of course, agents will use an internet financial system backed by digital assets, not banks and credit cards. Of course, agents will use cryptographic identities to identify, communicate, and transact with each other through decentralized protocols. I don’t see another solution: AI is on-chain. **