California AI regulation bill advances to Governor’s desk

A highly debated bill regulating artificial intelligence (AI) has cleared a significant hurdle in the California legislature. SB-1047, a bill that aims to regulate large AI models for safety, has been approved by the California State Assembly with 45 yes votes against 11 no votes. 

The bill, which was passed by the state Senate in May with a 32–1 vote, is expected to pass one more procedural vote in the Senate before it is sent for Governor Gavin Newsom’s signature. 

AI experts support bill to address potential AI threats

SB-1047 requires that developers of large-scale AI trained at a cost of over $100 million incorporate a ‘kill switch’ that shuts down the AI in case it develops adverse effects on the public. This approach is particularly relevant to cases when an AI system has to function with minimal human intervention, and worst-case scenarios may occur. 

The supporters of the bill’s passage claim that since advanced AI technologies are becoming more self-reliant, this bill is necessary to address the risks that come with it. The bill’s author and State Senator Scott Weiner pointed out that personalities like Geoffrey Hinton and Yoshua Bengio, leading experts in AI, expressed concerns about potential threats that AI may bring.

Last year, Hinton and Bengio were among the signatories of a statement that said if AI is not properly controlled, it could lead to human extinction. 

Critics voice concerns about Bill’s impact on innovation

However, SB-1047 has been met with a lot of criticism from supporters of AI research and business. Some have argued that this bill is more concerned with risks that may come in the future rather than concentrating on the current ones, such as deep fakes and fake news. 

Fei-Fei Li, the Director of the Stanford Institute of Human-Centered Artificial Intelligence, has been vocal against the bill. According to Li, in his recent editorial, the legislation is a threat to innovation because developers have a lot of exposure to liability. She noted that this could discourage the sharing of AI models and weights in the open-source format, which are essential for academics and collaboration.

California business leaders signed an open letter asking Newsom to veto the bill, which they said was “fundamentally flawed.” With SB-1047 heading towards enactment, Governor Gavin Newsom has until September 30 to approve the bill.

In May of this year, at a symposium at UC Berkeley, Governor Newsom agreed that proper regulation is a fine line. He raised concerns that the state could be put into a “very precarious situation” due to over-regulation. However, Newsom also recognized the necessity of regulation to some extent as the representatives of the AI industry raised certain questions.

Source: https://www.cryptopolitan.com/california-ai-bill-advance-to-governors-desk/