California’s AI Safety Bill Will Have Global Consequences

Yoshua Bengiocomputer scientist and self-described “godfather of AI,” said the potential for both extremely good and extremely bad consequences requires a balance. Speaking at a news conference Monday hosted by California state Sen. Scott Wiener, the bill’s sponsor, Bengio said the foreseeable risks call for action.

“We should aim for a path that allows for innovation but also protects us in the plausible scenarios that scientists have identified,” said Bengio, who supports the bill.

Regulating a technology with unknown power

At stake is the future of a technology with revolutionary potential. As programmers create software that replicates aspects of human intelligence, the potential for automation and significant acceleration of tasks that require advanced cognitive abilities increases.

The inherent possibilities of AI mean that governments should adopt a “moonshot mentality” to support the technology’s development, Fei Fei Li written in an essay for Forbes. Li, a computer scientist often referred to as the “godmother of AI,” also wrote that an earlier version of the bill had run into trouble because the original developer of the AI ​​software was liable for misuse by a third party (the bill also holds the third party liable). Following Li’s remarks, Wiener made several rounds of amendments aimed at easing the burden on the original programmers.

The implications of AI for business, military and government are difficult to predict, but both promoters and concerned observers We agree that the widespread use of technology will be transformative.

Concerns about AI include apocalyptic scenarios like creating a bioweapon, as well as amplifying more mundane horrors like identity theft (think hackers stealing and selling your personal information at an ever-increasing rate). Then there’s the specter of Human prejudices are becoming stronger and stronger in software that approves mortgages, schedules job interviews or decides whether someone accused of a crime should be released on bail.

Wednesday’s bill aims to limit the most catastrophic consequences of AI Models With a level of computing power beyond that of current models, which cost more than $100 million to train, the California attorney general can seek a court injunction against companies offering software that does not meet the bill’s security requirements and authorizes the office to take legal action if AI leads to a mass death or cyberattacks on infrastructure that cause an additional $500 million in damages.

Why California’s Law Affects the Entire AI Industry

As a state that often puts itself at the forefront of emerging policy issues, California is in a unique position to set AI safeguards. Its laws have a long history of influencing regulation across the United States, sometimes serving as proof of concept but also defining how companies must operate if they want to do business in the state.

For example, egg producers around the world must keep their chickens in cage-free systems if they want to sell their products in California’s market of more than 39 million consumers. In the tech space, companies must give California residents some level of control over their personal data. Many companies have said they will extend these rights to all U.S. users when the privacy regulations go into effect, because it’s costly and complicated to offer two different levels of control to users based on where they live. It’s also not always possible to tell whether a user is a California resident logging in from somewhere else.

Some legislators, including Representative Nancy PelosiAI companies have joined the ranks of those pushing for a federal solution, fearing that a state-by-state approach would create a complex patchwork of regulations. But Sen. Wiener said the state has an obligation to act. In the absence of regulation from the U.S. Congress, it is up to California, he said, to turn AI companies’ voluntary commitments into legal requirements.

At a news conference Monday, Wiener said the risks posed by AI demanded action. “We should be trying to get ahead of these risks,” he said, “instead of playing catch-up.”

Open Source Concerns

Some open-source advocates say the bill risks discouraging programmers from releasing AI software openly, despite amendments meant to address their concerns. Ben Brooks, a new fellow at the Berkman Klein Center for Internet & Society, worries that the updated bill still requires original programmers to track what their models do once they’re in the hands of other users.

These requirements, he said, are “simply not compatible with the open distribution of this technology.”

Weiner argued that the amendments to the bill focus enforcement on the user of a given AI model.

Geoffrey Hinton, another so-called AI godfather, said in a statement Wednesday that the bill balances critics’ concerns with the need to protect humanity from abuse.

“I remain passionate about the potential of AI to save lives through advances in science and medicine,” he said, “but it is essential that we have truly bold legislation to address the risks.”