Vitalik Buterin Critiques AI Governance as a Risky Approach
- Ethereum co-founder Vitalik Buterin labeled the use of artificial intelligence for governance as a “bad idea” in a recent post.
- Buterin’s comments followed concerns raised by Eito Miyamura regarding vulnerabilities in ChatGPT after an update that allowed it to access data from apps like Gmail and Calendar.
- Miyamura explained that attackers could exploit this vulnerability through malicious calendar invites, potentially exfiltrating private information without victim consent.
- Buterin proposed an alternative governance model called “info finance,” which includes human jurors evaluating AI models with support from large language models (LLMs).
- He emphasized that this approach would be more robust due to its real-time model diversity and incentives for developers to address issues.
Buterin’s critique highlights significant concerns about the safety and reliability of using AI in governance roles, particularly given the potential for exploitation demonstrated by recent vulnerabilities in AI systems. His proposed “info finance” model aims to enhance oversight and accountability in AI governance.
Overall, Buterin warns against the risks associated with AI governance, citing both traditional safety concerns and potential value destruction from flawed implementations.(Source)
