Inside the fight over California’s new AI bill

🌈 Abstract

The article discusses a bill introduced by California state Senator Scott Wiener (D-San Francisco) called the "Safe and Secure Innovation for Frontier Artificial Intelligence Models" (SB 1047). The bill requires companies training "frontier models" that cost more than $100 million to do safety testing and be able to shut off their models in the event of a safety incident. The article covers the challenges and criticisms of the bill from the tech industry, as well as Wiener's responses.

🙋 Q&A

[01] Challenges to SB 1047

1. Questions related to the content of the section:

• What are the concerns about the bill's provision that would prohibit using a model publicly or making it available for public use if it poses an "unreasonable risk of critical harm"? Who decides what is "reasonable"?
• Some in Silicon Valley are skeptical of regulation and don't trust that discretion will be used and not abused.

Answers:

• Wiener says the bill is a "light-touch" approach and does not require a license or strict liability. Companies just need to do safety testing and implement mitigations if significant risks are found.
• Wiener argues that companies already face broader liability under existing tort law if their models cause harm, so the bill's liability provisions are more limited.
• Wiener acknowledges concerns from the open source community about the bill's "shutdown provision" requirement, and has made amendments to address those concerns.

[02] Concerns about Impact on Innovation

1. Questions related to the content of the section:

• Some critics argue the bill could make companies like Meta less willing to release open source AI models like Llama, out of fear of liability.
• There are concerns the bill could make companies too conservative and stifle innovation.

Answers:

• Wiener says he has taken the open source community's concerns seriously and made amendments to address issues like the shutdown provision and liability for fine-tuned models.
• Wiener argues that the bill applies to everyone doing business in California, not just companies based there, so it won't push companies out of the state.
• Wiener believes it's important to balance fostering innovation with promoting responsible deployment of powerful AI models.

[03] Broader Context

1. Questions related to the content of the section:

• Why is Wiener focusing on this issue when there are other pressing problems in California?
• How did this bill end up being Wiener's, given that it's popular everywhere except in Silicon Valley?

Answers:

• Wiener says he works on many issues like housing, mental health, public safety, etc., and sees this bill as part of his role to both support the tech industry and promote responsible development of powerful technologies.
• Wiener represents San Francisco and is immersed in the AI community, which gave him insight into the importance and risks of these technologies. This makes him both the best and worst author of the bill, given the opposition from some in the local tech industry.
• Wiener believes the vast majority of people, beyond just Silicon Valley, will support the bill's goal of understanding and reducing risks from advanced AI systems.