原文标题：AI-powered smart contracts could be ‘transformative’ — Ava Labs founder
原文作者：Andrew Singer
原文来源：Cointelegraph
Compilation: Than push an
Imagine if the average person could write smart contracts in their native language, it could bring "billions of new blockchain users."
Emin Gün Sire, founder of Ava Labs and the one who built the Avalanche blockchain, said, "Smart contract programming is hard. Because capturing "intent" is hard, writing code is tricky, and validation is still a challenge."
Speaking at the recent Cornell Blockchain Conference in New York City, Sirer said, "This is a major issue that is holding back smart contracts from thriving. “
But imagine if an artificial intelligence (AI) like ChatGPT or Llama could do all the smart contract coding for you. Lawyers may soon become the primary writers of smart contracts as opposed to programmers.
What's even more interesting is that ordinary people will be able to write smart contracts in a peer-to-peer manner in their native language, which will become as easy as writing a bank check.
At the Cornell Blockchain Conference in New York City on April 26, Sirir further elaborated on this vision for the future:
"You're all familiar with the process of writing a check, right? That's the most sophisticated financial instrument our bank offers. At the moment, our options are rather limited. For example, I could write a check to 'John' that said, 'I hereby transfer $5,000 to John' – that's pretty much it. ”
But if we can do the same thing, and be able to add additional conditions, Sirer goes on to illustrate:
"I'm offering $5,000 to John on the condition that he has raised $5,000,000 to make a film by the end of September. If he can't, I want my money back. I can say it directly. ”
Mr Sirir said the scenario he envisioned could take another five to 10 years to materialize, but if it did, it would be "transformative".
"We'll be able to onboard billions of new [blockchain] users."
Currently, most smart contracts are written using Solidity, a programming language that is not well known even among programmers.
However, Avalanche is already working on a new type of virtual machine in the field of artificial intelligence and blockchain technology. According to Sirer, the virtual machine "enables programming in natural language." You can choose to write your program in English, German, French, Tagalog, Chinese, etc., which is any natural language your mother taught you in your mother tongue. ”
The problem still exists
But is this just a pipe dream? There are still a number of issues that need to be addressed before what Sirer calls "coin-operated brokers" can be widely adopted.
First of all, it is necessary to deal with the legal aspects. Keywords and terms need to be precisely defined. For example, the term "software" may need to be defined in great detail before it can be used in a smart contract. It may take some time to complete these tasks.
In addition, it remains to be seen whether the AI "agents" are ready to write legally binding documents. After all, large language models (LLMs) sometimes produce false information, a phenomenon sometimes referred to as "hallucinations."
"You've definitely seen [AI-generated] images of people with seven fingers," Sirer notes.
However, some of these problems can be solved in the short term. For example, transactions can be submitted to both ChatGPT 3.5 and Llama, requiring both agents to agree before a smart contract solution can be adopted. "That way, at least you've got two trusted executors, and you're taking their actions at the intersection."
"Instructions for writing smart contracts in plain and concise English"
It has also been suggested that smart contracts capable of handling "natural" languages are already available – or arguably within reach.
Sam Friedman, Chief Solutions Architect at Chainlink Labs, told Cointelegraph, "In today's world, the ability for everyone to be their own programmer already exists. ”
Friedman mentioned one of the projects in Chainlink's BUILD initiative, the Council, an AI proxy framework developed by ChainML that "allows developers to provide simple English instructions and instantly translate them into executable code." Currently, this code is mostly Python or SQL, but to implement the functionality of a smart contract, it only takes one upgrade. ”
He also mentioned the Chainlink Developer Center's "Ask AI" feature, which returns a full smart contract when a user submits a request in plain English. For example, a user can request something like this: "Please show me an example of a smart contract that allows the user to request a random number between 1 and 10 and should use Chainlink VRF (Verifiable Random Function)." ”
However, the current product may not fully implement all the features envisioned by Sirer, and Friedman said in an interview with Cointelegraph that there is still a lot of work ahead.
"Although AI agents who write smart contracts have emerged, they have not yet reached their final form," he said. They still need further improvement in terms of quality and complexity. "In order to achieve this, AI models need to be trained with content that is constantly updated.
Even so, Friedman was able to imagine use cases where multiple AI "agents" with different roles worked together to complete the lifecycle management of smart contracts. For example:
An agent helps define requirements;
Another agent is responsible for writing smart contracts;
There is also an agent that performs quality and safety testing;
Another agent is responsible for deploying the contract;
The last agent performs ongoing maintenance or adjustments to the parameters or global variables in the contract after the contract is deployed.
At the Cornell Blockchain Conference, Sirer acknowledged that there are still "huge unknowns" and that progress will require guidance from all sides, as well as the participation of "people with both technical backgrounds, as well as ethical and human backgrounds."
There is also a place for lawyers?
During the Q&A session that followed the presentation, Sirer was asked who would write smart contracts in the future if they no longer needed to use the Solidity language and could be coded in ASCII text.
So who would be the ideal programmer? At first glance, software programmers seem to have the upper hand. After all, they have been able to use precise language to define and specify conditions proficiently.
"But it's the lawyers who really know how to use normative language to make contracts. As a result, they will become leaders in that field[......] are the ones who can really thrive in this system. ”
What about the average user? "I think they're perfectly capable of defining a fundamental deal," says Sirer. In fact, he expects the new features to be built on "democratic, open platforms" that "can be used equally by anyone."
Overall, this new approach to smart contracts is very appealing and promising, "and we can't just put it on the back burner," Sirer concluded.