AI or Not CEO's thoughts on California Governor Gavin Newsom vetoing the AI safety bill SB 1047.
CEO @ AI or Not | Protecting companies against fraud, deepfakes, misinformation and KYC scams coming from generative AI.
Good on California Governor Gavin Newsom for vetoing the AI safety bill SB 1047. Though its intent was for AI safety, it missed the mark on some key points. SB 1047 summary: ? Required developers of large AI models, ie LLMs, to take “reasonable care” to ensure that their technology didn’t pose an “unreasonable risk of causing or materially enabling a critical harm.” ? Critical harm was defined as mass casualties or over $500 million in damages. ? Large AI models were defined as costing over $100 million to develop or $10 million to fine-tune. ? Each model was to include a kill switch to turn off the AI by a human immediately, if needed. Where the bill misses: ? Targeting only large AI models will force companies to either i) play accounting games to stay under that $100m amount or ii) prohibit funding, both venture and corporate R&D, thus hampering innovation for future models in order to avoid government scrutiny ? In practice, fine-tuning an existing, already in the wild, open source model costs a lot closer to $20 (no $m!) than it does $10m. And this is already happening now. ? Mass casualties and $500m in damages are quite extreme; consider the CrowdStrike outage cost Delta Airlines around that so its enough to cripple large companies and wipe out all others. Many technologists were celebrating this outcome, including Marc Andreesen and Andrew Ng. But there were supporters for the bill too, including Elon Musk and Anthropic's CEO Dario Amodei. I understand why they took they showed support but still think the ambiguity and prohibiting areas did more harm than good. Back to building!
