The Endgame of California's Anti-AI Bill
The Endgame of California's Anti-AI Bill
Fool me once, shame on you. Fool me twice, shame on me.
SB 1047 was amended. It has genuinely improved in some ways, most notably when it comes to criminal liability. Instead of arresting researchers for publishing open source models, which Senator Wiener initially denied the bill was doing, it only fines them enough to bankrupt any individual or small company. But that’s better than being thrown in jail.
However, the main problem – fining and punishing AI research instead of bad actors remains. To quote U.S. House Science Committee Ranking Member Zoe Lofgren (D-CA18):
The problematic core concerns remain: there is little evidentiary basis for the bill; the bill would negatively affect open-source development by applying liability to downstream use; it uses arbitrary thresholds not backed in science; and, catastrophic risk activities, like nuclear or biological deterrence, should be conducted at a federal level.
Let me make it explicit about what she has implied far more carefully. This bill continues to harm AI research, particularly open source, in order to alleviate the baseless, non-empirical concerns of unscientific Doomer billionaires who believe, verbatim, in “human extinction”, paying the salaries of bill co-sponsor Center for AI Safety.
SB 1047 has gone from the equivalent of nuking silicon valley thrice, to only nuking it twice. These changes should have come in April. They would have come in April if Wiener was trying to make genuine attempts to improve the bill.
Fool me once, shame on you. Fool me twice, shame on me. Let me explain why they came in August in terms founders will understand.
Companies have momentum. With their product launch, they want to get hype going and define the narrative. Legislators are no different. By delaying these basic and insufficient changes to August, Scott Wiener is intentionally attempting to change the narrative while maximally continuing to promote a bill that serves only AI Doom’s billionaire donors.
One change that makes this deception clear is the renaming of the “Frontier Model Division” to the “Board of Frontier Models”. The co-sponsor of the bill, the Doomer-backed Center for AI Safety, advertised this as “Elimination of the Frontier Model Division”. This is a clearly misleading headline; the Frontier Model Division was eliminated in name only.
This comes after intense opposition from all corners of America. Elected Democrats oppose the bill, including former House Speaker Nancy Pelosi (D-CA11), U.S. House Science Committee Democrats, and former interim mayor Mark Farrell. Mainstream academics have also voiced their opposition, including Fei-Fei Li, the “Godmother of AI”, Ion Stoica, director of the University of California, Berkeley's Sky Computing Lab, and Andrew Ng, Coursera co-founder and Stanford professor. Y Combinator and a16z sent open letters with hundreds of startups each opposing the bill.
SB 1047 is a microcosm of Doomer influence. They are using dirty tricks that hurt their reputation in the long term because they don’t think they have long. This is either because they think (incorrectly) that Doomsday is imminent, or because they think (correctly) that the views of politicians and the public are rapidly moving against them.
For a full list of changes in SB 1047, I recommend Dean Ball’s excellent article:
