In a brand new letter, OpenAI chief technique officer Jason Kwon insists that AI rules must be left to the federal authorities. As reported beforehand by Bloomberg, Kwon says {that a} new AI security invoice into account in California may sluggish progress and trigger firms to depart the state.
A federally-driven set of AI insurance policies, reasonably than a patchwork of state legal guidelines, will foster innovation and place the U.S. to steer the event of worldwide requirements. Because of this, we be a part of different AI labs, builders, specialists and members of California’s Congressional delegation in respectfully opposing SB 1047 and welcome the chance to stipulate a few of our key considerations.
The letter is addressed to California State Senator Scott Wiener, who initially launched SB 1047, often known as the Secure and Safe Innovation for Frontier Synthetic Intelligence Fashions Act.
Based on proponents like Wiener, it establishes requirements forward of the event of extra highly effective AI fashions, requires precautions like pre-deployment security testing and different safeguards, provides whistleblower protections for workers of AI labs, offers California’s Lawyer Normal energy to take authorized motion if AI fashions trigger hurt, and requires establishing a “public cloud pc cluster” referred to as CalCompute.
In a response to the letter printed Wednesday night, Wiener factors out that the proposed necessities apply to any firm doing enterprise in California, whether or not they’re headquartered within the state or not, so the argument “is not sensible.” He additionally writes that OpenAI “…doesn’t criticize a single provision of the invoice” and closes by saying, “SB 1047 is a extremely affordable invoice that asks massive AI labs to do what they’ve already dedicated to doing, specifically, check their massive fashions for catastrophic security threat.”
The invoice is at the moment awaiting its closing vote earlier than going to Governor Gavin Newsom’s desk.
Right here is OpenAI’s letter in full: