Vitalik warns devs to ‘be careful’ when adding butter to blockchains
Ethereum founder Vitalik Buterin warns developers to “be careful” when integrating artificial intelligence (AI) with blockchain technology.
In a Jan. 30 blog post, Buterin shared his biggest concerns with AI, telling developers to “tread carefully” especially when deploying new apps in high-value, high-risk contexts.
It's important to be careful: if someone builds a stablecoin that uses, for example, a prediction market or an AI oracle, and the oracle gets hacked, that's a huge amount of money that can be lost in an instant.
Buterin explained that one area where AI has the greatest “potential” is allowing AI to engage in micro-scale techniques that are not suitable for humans, such as prediction markets.
AI can also be used in crypto wallets for users, which can help explain transactions and signatures or detect fraud and securely interact with crypto, Buterin wrote.
“My opinion is that pure AI interfaces are probably the most dangerous at the moment, because they increase the risk of other types of errors, but AI is becoming more and more effective in meeting a more common interface,” he said.
The promise and challenges of crypto + AI applications https://t.co/ds9mLnshLU
— vitalik.eth (@VitalikButerin) January 30, 2024
In Buterin's view, the most dangerous endeavor is enforcing the rules or governance of crypto systems using AI, as open source AI models are vulnerable to adversarial attacks.
Open source AI is more vulnerable to adversarial attacks because malicious actors can inspect its code and modify attacks against it. Closed source AI provides “security in disguise”, but these models do not provide any clarity on how they work or guarantee that they are not biased.
An example of closed-source AI from OpenAI-adjacent crypto startup Worldcoin: Buterin explains that it relies on “not allowing anyone to easily call the AI model” and is instead using “trusted hardware.”
Related: Can blockchain provide safeguards to keep AI on course?
Buterin explained that the biggest challenge is creating decentralized AI using crypto and blockchain that other applications can use.
The most challenging applications to be precise are applications that attempt to use blockchains and cryptographic techniques to create a “single” decentralized, trusted AI that some applications rely on for a specific purpose.
These applications promise to “improve both the practicality and security of AI in a way that avoids the central risks associated with mainstream approaches to that problem.” There are also many ways in which the underlying assumptions can fail, he said.
Magazine: Real AI use cases in crypto, No. 1: The best currency for AI is crypto.
