California is at the forefront of AI regulation with the introduction of Senate Bill 1047 (SB 1047), which aims to prevent potential disasters caused by advanced AI systems. Spearheaded by Senator Scott Wiener, the bill is designed to impose stringent safety protocols on developers of large AI models. However, it has sparked significant controversy among Silicon Valley giants and AI startups alike.
What SB 1047 Proposes
SB 1047 focuses on preventing “critical harms” that could be inflicted by AI systems, such as mass casualties or extensive cyberattacks. The bill would apply to the most powerful AI models, those requiring $100 million or more to train and utilizing 10^26 FLOPS. It mandates the implementation of rigorous safety protocols, including an “emergency stop” feature and annual third-party audits. A new state agency, the Frontier Model Division (FMD), would oversee compliance.
Proponents’ Perspective
Supporters of SB 1047, including Senator Wiener, argue that the bill is a proactive measure to prevent AI-related disasters. The bill has garnered support from prominent AI researchers, such as Geoffrey Hinton and Yoshua Bengio, who are concerned about the existential risks posed by unchecked AI development. The Center for AI Safety also backs the bill, arguing that it could prevent a major AI incident that could stifle future innovation.
Opposition from Silicon Valley
Silicon Valley, however, has voiced strong opposition to SB 1047. Critics, including venture capital firm Andreessen Horowitz (a16z), argue that the bill’s thresholds are arbitrary and could stifle innovation by imposing onerous requirements on startups as AI technology becomes more expensive. High-profile AI researchers, such as Fei-Fei Li and Andrew Ng, have also expressed concerns that the bill could harm the open-source AI community and hinder AI research.
The Path Forward
Despite the pushback, SB 1047 is expected to pass in the California Senate, though not without amendments. AI company Anthropic has suggested several changes to the bill, including reducing the power of the Attorney General to sue developers and removing whistleblower protections. These amendments are currently under consideration.
If passed, SB 1047 will set a significant precedent for AI regulation, though its impact on the tech industry remains hotly debated. Governor Gavin Newsom will have the final say, with the bill’s fate to be decided by the end of August 2024.
