Movies

AI safety bill passes California Legislature

BBG News

AI safety bill passes California Legislature

As a seasoned tech journalist with over two decades of covering the Silicon Valley scene, I’ve seen my fair share of groundbreaking innovations and controversial bills that have shaken the industry to its core. The latest bill on the block, SB 1047, is no exception.

A contentious piece of legislation, aimed at making developers implement safety precautions for sophisticated AI systems, moves a step nearer to being enacted as law.

In simpler terms, Bill SB 1047 proposes that creators of innovative AI systems in the future should establish safeguards to stop these technologies from being exploited for launching cyber-attacks on essential structures like power stations.

Developers must provide their safety protocols to the Attorney General. If any AI models they personally manage lead to harm or an immediate risk to public safety, they could be held accountable for these actions.

The legislation proposed by Senator Scott Wiener from San Francisco’s Democratic party was endorsed by the State Assembly on Wednesday, garnering 41 favorable votes against nine oppositions. On Thursday, this proposal received approval in the State Senate through a concurrence vote. At this point, it is moving towards Governor Gavin Newsom’s office; however, it remains undecided whether he will sign or veto the bill.

“Innovation and safety can be combined effectively, and California is setting an example for this,” Wiener pointed out in his statement.

A spokesperson for Newsom said the governor will evaluate the bill when it reaches his desk.

In the vibrant tech scene of the Bay Area, Wiener’s bill sparked intense discussions. The Center for AI Safety, Elon Musk (CEO of Tesla), the L.A. Times editorial board, and an AI startup based in San Francisco called Anthropic all voiced their support for it.

However, the bill faced opposition not only from Democratic congressional leaders but also tech giants like Meta and OpenAI, who voiced apprehensions that it might curtail technological advancements in artificial intelligence within California.

Leading Democrats in the U.S. House of Representatives such as former Speaker Nancy Pelosi, Rep. Ro Khanna from Fremont, and Zoe Lofgren from San Jose, have voiced their opposition to the bill and advised Governor Newsom to reject it. They contend that the law could potentially harm California’s flourishing AI sector, which hosts companies like OpenAI, the creator of ChatGPT. Furthermore, they mention ongoing congressional initiatives focusing on AI as a reason for their stance.

“In a letter to Governor Newsom, a group of seven Democratic congressional representatives expressed their concern that businesses might choose to establish themselves in different countries or refrain from introducing their models in California, due to a significant potential risk.”

As a film buff passionate about the intersection of technology and cinema, I strongly believe that California should spearhead advancements in AI while safeguarding consumer rights, data privacy, intellectual property, and other crucial aspects. However, in my view, Senate Bill 1047 unfortunately doesn’t aid this mission; instead, it seems to cause more harm than good in our pursuit of responsible AI development.

As a movie critic, I’d say this legislation, opposed by Wiener and others, is touted as a double-edged sword. They argue it encourages innovation, while at the same time safeguarding the public interest.

“Before the Assembly’s vote on Wednesday afternoon, Assemblymember Devon Mathis (R-Visalia) suggested that guardrails need to be installed. He emphasized the importance of ensuring that everyone involved is reliable.”

Advocates of SB 1047 argue that it places the duty of ensuring safety on developers who use sophisticated AI systems. This measure could potentially stop harmful AI incidents from occurring in the long run.

At a recent press conference, Yoshua Bengio, a professor at Université de Montréal and the head of Mila – Quebec Artificial Intelligence Institute, expressed concern that tech firms may not address the major risks linked to AI independently. This is because they’re so focused on competing for market share and maximizing profits that they might not be able to ensure safety effectively. He likened this situation to students grading their own homework, suggesting instead that we cannot rely solely on them and should take proactive measures to mitigate these risks.

Supporters argue that, much like other sectors with possible safety concerns, AI should be governed similarly.

In simpler terms, Musk expressed his opinion on a platform on Monday, stating that California should likely approve SB 1047, an AI safety bill. He’s been advocating for regulations on artificial intelligence for over two decades, just as we regulate any product or technology that could potentially harm the public.

In the past few days, a bill underwent an essential evaluation in a state Senate panel following substantial amendments by Wiener. Among these alterations were the deletion of a penalty for perjury and adjusting the legal criteria that developers must meet concerning the safety of their sophisticated AI systems.

In a letter dated August 21st, Anthropic’s CEO Dario Amodei, based in San Francisco, expressed his belief that while there may be costs associated with the bill, its benefits are likely greater. However, it’s important to note that this letter did not offer an endorsement of the bill, but rather presented Anthropic’s perspective on its potential advantages and drawbacks.

“Amodei made it plain that SB 1047 aims to tackle genuine and significant issues related to catastrophic risks in artificial intelligence systems. The rapid advancement of AI technology presents a double-edged sword, offering immense potential for California’s economy while also posing substantial risks.”

But some tech companies like OpenAI said they opposed the bill even after the changes.

In a letter dated August 21st, OpenAI’s Chief Strategy Officer Jason Kwon emphasized that due to the far-reaching and substantial impact of AI on U.S. competitiveness and security, it is crucial for federal authorities to establish and enforce regulations regarding advanced AI models. He argued that a uniform set of AI policies at the federal level would promote innovation and place the U.S. in a leading position to shape global standards, as opposed to a patchwork of state laws.

Wiener said he would welcome a strong federal AI safety law that preempts his bill.

“Based on past events, passing this federal law could prove challenging,” Wiener stated. “However, until then, California should persist in advocating for legislation such as SB 1047 that encourages innovation and safeguards the public.”

Among around fifty legislative proposals concerning artificial intelligence, SB 1047 is one of them, focusing on different dimensions of how this technology affects society, such as employment, deepfakes, and general safety, within the state’s Legislature.

Read More

2024-08-29 23:31

Previous post Outer Banks season 4: Two release dates, teaser, and feast your eyes on the cute (and not so cute) OBX couples!