Rep. Nancy Pelosi has come out in opposition to a high-profile California state invoice that might regulate Huge Tech’s fast-growing synthetic intelligence know-how – describing the invoice as “well-intentioned however ill-informed.”
California’s SB 1047 would set up security requirements for AI fashions that value greater than $100 million to coach and mandate pre-release security testing for so-called “frontier” AI fashions, amongst different necessities. It’s opposed by the tech business, together with Meta, Google and enterprise agency Andreesen Horowitz.
Pelosi (D-Calif.) – who has come underneath hearth for owning a lucrative stock portfolio that features AI chip provider Nvidia, software program agency Databricks and different main tech corporations whereas nonetheless in workplace – argued the invoice may stifle innovation.
“Whereas we wish California to guide in AI in a manner that protects shoppers, knowledge, mental property and extra, SB 1047 is extra dangerous than useful in that pursuit,” Pelosi stated in a prolonged assertion.
Pelosi famous that a number of California Democratic lawmakers, together with Reps. Zoe Lofgren, Anna Eshoo and Ro Khanna and Gov. Gavin Newsom, have expressed issues in regards to the laws or outright opposed it.
She additionally cited the view of Stanford College AI scholar Fei-Fei Li, who has warned the invoice “would have important unintended penalties that might stifle innovation and can hurt the US AI ecosystem,” based on Pelosi’s assertion.
The invoice was launched by California State Sen. Scott Weiner, who’s broadly anticipated to run in opposition to Pelosi’s daughter, Christine, to take over her US Home seat when the previous speaker leaves workplace.
Weiner stated he “respectfully and strongly” disagreed with Pelosi.
“Innovation and security should not mutually unique, and I reject the false declare that to be able to innovate, we should depart security solely within the palms of know-how firms and enterprise capitalists,” Weiner stated.
The laws is a key combat within the nationwide debate over how the government should regulate the powerful AI models constructed by the likes of Google and Sam Altman’s OpenAI.
Critics warn that superior AI techniques could pose a major risk to society – from the unfold of misinformation to even the destruction of humanity – with out correct guardrails in place.
SB 1047 set to obtain a vote within the California state meeting by the top of August.
The invoice’s detractors assert that the laws will solely damage the US in its bid to guide the race of superior AI.
In June, Meta AI chief Yann LeCun argued the bill’s approach to AI regulation “will put an finish to innovation.”
In an effort to assuage the business’s issues, California lawmakers made adjustments to the invoice earlier this month, resembling dropping plans to create a authorities company known as the Frontier Mannequin Division (FMD) to supervise the security overhaul.
Lawmakers additionally dropped a clause that might have allowed California’s state legal professional basic to straight sue AI corporations in the event that they had been decided to be negligent their security practices.
Anthropic, the AI large backed by Google and Amazon, had pushed for that change. After the adjustments had been introduced, Anthropic told TechCrunch that it was nonetheless reviewing the amended invoice and that not all of its proposals had been adopted.
Nonetheless, the amended model of the invoice hasn’t glad all detractors.
“The edits are window dressing,” Andreesen Horowitz accomplice Martin Casado wrote on X. “They don’t tackle the actual points or criticisms of the invoice.”