Ethereum co-founder Vitalik Buterin cautioned developers about the risks and complexities involved in integrating artificial intelligence (AI) with blockchain technology. Highlighting the need for careful consideration, especially in high-risk and high-value applications, Buterin pointed out the potential dangers and opportunities of this fusion.
Buterin stressed the vulnerability of AI oracles in critical blockchain applications, such as prediction markets or stablecoins. He warned that if these AI systems are compromised, there could be significant financial repercussions. However, he also recognized the beneficial role AI can play in blockchain technology. AI’s potential to facilitate micro-scale operations in prediction markets, which are otherwise impractical for humans, and its ability to enhance user experience in crypto wallets through transaction guidance and scam detection are notable.
Buterin advised a cautious approach towards AI interfaces, suggesting that while they are risky due to the potential for errors, AI can effectively augment traditional interfaces. He particularly highlighted the risks associated with using AI in the governance of crypto systems, noting that open-source AI models are prone to adversarial attacks. In contrast, closed-source AI offers more security but lacks transparency and assurance of unbiased operations.
Using Worldcoin, a crypto startup related to OpenAI, as an example, Buterin explained how closed-source AI implementations, which use trusted hardware to control access to their AI models, can offer a balance between security and functionality.
Buterin also delved into the challenge of creating a decentralized AI using blockchain technology. This concept, referred to as a “singleton,” involves developing a single, decentralized, trusted AI for diverse applications. He emphasized the potential of such developments in enhancing functionality and AI safety, and in avoiding the centralization risks associated with mainstream approaches.
However, Buterin concluded by acknowledging the potential risks and the possibility of failure in the underlying assumptions of these innovative applications. His insights underscore the need for caution and thorough assessment in the ongoing exploration of AI and blockchain integration.
