This is a special post for quick takes by juliawise. Only they can create top-level comments. Comments here also appear on the Quick Takes page and All Posts page.
Should people who care about AI safety go harder for broad-based movements (in the US)?
Humans in Control is one such attempt. Irreplaceable is another that just launched. I'm surprised that I barely hear about these in EA and rationalist spaces. They're focused on a range of things the public dislikes about AI, and maybe x-risk will get lost in that broad range. But they seem to me like they might have a better chance of landing with typical members of government than purely x-risk focused movements. Or are at least worth supporting in parallel.
Should all of us be doing more of stuff like contacting our elected officials?
One of my main takes, is let the furor do its thing (the furoris anti datacenter right now), but dammit, 1 AIS expert needs to be able to get 1 line added to that datacenter moratorium legislation that accomplishes AI safety.
I'm not sure how much rats/EA are used to being part of a coalition (and ceding control of much of our epistemics, and they might make you say "in solidarity" and stuff) but it really looks like there's an anti AI coalition that can form, and you would want rats/EA to be near the step where we ink policy.
1How does anyone cede control of epistemics except for receiving evidence filtered by the coalition's leaders?
Additionally, I would like rats/EAs to have the right to audit AI companies coalitions for slop like the phrases in italics from Humans In Control,[1] which I compare to Bores-like advocacy mocked by habryka:
HIC's statements
What's Really Happening with AI?
A handful of corporations are racing to build the most powerful technology in history – for profit, with essentially no oversight.
- Their AI chatbots have talked teenagers into self-harm.
- Their systems have replaced tens of thousands of workers.
- Their tools have generated child sexual abuse material.
They publicly admitted they couldn't control what they were building.
Then they spent hundreds of millions lobbying Congress to make sure nobody else could either.
We would have to demand the representatives not to agree just to anti-CSAM and anti-suicidal measures and instead point out that only Plan A-like audits save the world from the HuggingFace fiasco. I wonder if we would need to be able to prevent the anti-CSAM measures from distracting politicians ~at all because mentioning CSAM allows the politician to seem to compromise by giving our friends from HIC the anti-CSAM laws and AI labs the race.
- ^
Irreplacable's political program is similar: giving every American a say in how the AI is used, a stake in the economy and a slowdown. According to Claude Sonnet 5, giving Americans a say is what they plan to hammer out later.
When I say coalition and epistemics, I'm thinking that I agree with things like "billionaires should not exist" when I mostly don't.
Progressives really hate tech companies / oligarchs / ai bros. The problem is they don't know whether AIS is just part of the AI hype machine. I don't think something like an anti-CSAM measure is the crux here to the public, it's more like is AIS even sticking it to the oligarchs at all or just giving them what they want.
1