This is a special post for quick takes by Noah Rossi. Only they can create top-level comments. Comments here also appear on the Quick Takes page and All Posts page.
Takeaways from calling my congressional representatives about AI risk
Two days ago, I called my two U.S. senators and my representative to express my concerns about AI risk.[1] For both my senators, I got sent to voicemail, and I haven't heard anything since then. For my representative, a staffer answered, and I outlined my concerns about loss of control. I also urged the congresswoman to pass the CATS Act and AI Incident Reporting Act.
I was very concerned when I received this follow-up email (emphasis mine):
[...]
So far this decade, we as humans have seen the development and advancement of AI reach heights never thought possible. While AI can be a useful tool for everything from screening for diabetes to acquiring information on a given subject, it can also pose dangers to the general public. That’s why, in Congress, I have raised red flags about AI chatbots and their role in mental health support for America’s youth. To that end, I believe that AI needs to be a balanced and refined tool that is used for good purposes and for the betterment of society.
Legislation that oversees AI-related policy is handled by the House Committee on Energy and Commerce. As a member of this committee, I would be happy to keep your thoughts on AI in mind should it come before me for a vote on the House floor.
[...]
I'm most worried that this email suggests that AI risks come from "chatbots," rather than agents or agent swarms, and that it only names youth mental health as a risk, not loss of control or misuse of powerful AI systems. After the Hugging Face incident, I would have expected that congressional offices would be better informed about AI risk, but this responses updates me against that.
To steelman my representative (Rep. Miller-Meeks): she already has co-sponsored the CATS Act. It's also possible that the follow-up I received was a template that hasn't been updated since last year. Also, my representatives (Iowa's 1st district) might be less informed about AI risk than other representatives, given the lack of AI labs or application-layer startups in the state.
I used callcongress.ai, which made this fast and easy.
If it’s been a few weeks I’d say call again especially with things changing fast + not having actually gotten a hold of anyone. I’ve gotten the boilerplate non-xrisk-addressing emails a couple of times after contacting reps. However I got a startled/taken aback sort of response on the phone from one of my reps’ staffers when I called a couple days post-the Coxon story blowing up. I need to get going on actually requesting a meeting. It’s a heavy topic with some counterintuitive abstract parts and there’s no substitute for actually being face to face with s... (read more)