The Tech Industry is the Biggest Blocker to Meaningful AI Safety Regulations

In 2022 and early 2023, I remember having some conversations with folks about whether or not industry groups would support regulation. (At the time, the thought of governments becoming concerned enough about risks from advanced AI to consider regulation was one of those things that seemed like it wouldn't happen for a long time, if ever). 

Perhaps I and some others were naive. But I genuinely bought the idea that if governments ended up getting more concerned about AI risks, then some of the industry players would end up supporting common-sense regulations. Especially industry players that had acknowledged AI risks (e.g., Anthropic, OpenAI, DeepMind). I think this was also a fairly commonly-held opinion within AI safety circles (and indeed was part of the story for why people concerned about AI safety should support Anthropic, OpenAI, and DeepMind).

In some ways, it's very sad to see how wrong these takes were. The "maybe industry will support regulations or maybe even proactively push for regulations" takes have not aged well.

In other ways, I think the SB1047 saga has mobilized a lot of AI governance/policy folks and caused them to realize where industry stands. There is a bit of poetic Litany of Gendlin [? · GW] energy here– if industry is not going to support even light-touch regulation, it is worth swallowing this difficult pill and orienting to the landscape accordingly.

(That said, I am interested in seeing how Anthropic reacts to the amendments. I think Anthropic has done a lot less for comms/policy efforts than many folks had predicted it would, and I think many are uncertain if its comms/policy efforts will differ meaningfully from its competitors.) 

(I also think that as evidence of risk increases, some industry players might either change their minds or feel compelled to support regulation out of fear of looking super unreasonable.)