tl;dr, GOP presidential candidate Will Hurd seems to be making AI alignment a key part of his platform. Is it worth trying to help him to get onto the debate stage?

disclaimers: 1. this post is about politics, obviously; 2. although I am a director at Cavendish Labs, everything expressed in this post is done entirely in a personal capacity, and in no way reflects any opinions of Cavendish Labs.

epistemic status: highly uncertain. mostly quick thoughts on my impressions, plus ten minutes of research. written in like 5 minutes.
 

So I was driving through New Hampshire today (on the way from Boston to Cavendish), when suddenly a thought hit me—aren’t people campaigning for president around here? So I pulled up some GOP events calendar, and indeed, Will Hurd’s event was starting in 20 minutes, a 7 minute drive away. I’d heard of Will Hurd before—but only in the context of him being a candidate polling at 0%. I went on his website, and came out pretty unimpressed; it seemed like a platform of a generic also-ran that might be fun to stumble upon on archive.org in 2026 and say “wow, totally forgot about this guy!”. But anyways, the allure of meeting a presidential candidate drew me in, and I pulled up to the St. Anselm’s College Institute of Politics.

There were probably around 15 people there, including his crew, an NBC camerawoman, and a smattering of 65-year-olds who seemed to be experienced political event-goers (two of them had been at a different candidate’s event just last night!) Hurd does the classic politician thing of walking around, shaking everyone’s hands, and flattering people (“you look familiar. have we met before? no? perhaps you’ve starred in a movie?”), and then goes on to start his stump speech. He recounts a few CIA stories, highlights his “Granite State” bona fides, and then—shockingly—starts talking about alignment. He uses the word, and everything. He gives a nice short explanation for the audience, which is basically “the larger AI systems become, the worse we understand what they do, and how to make them do what we want”—and talks about his policy plan, which is basically requiring licenses to train or operate AI models past a certain size or powerfulness.

I talked to him one-on-one for about ten minutes afterwards—he clearly knows his shit, and seems quite talented at finding good people to talk to about AI policy (also, he either currently is, or was, on the board of OpenAI). Anyways, I’m pretty thoroughly impressed; I was not expecting this calibre of reasoning from a “major” GOP presidential candidate.

Some thoughts and questions:

1. so, he has a 0% chance of winning the primary. are there any other serious, potentially highly impactful politicians who seem to care about alignment?

2. would it make any sense to try to get him onto the debate stage? given that i) he brought up alignment in a random 13-person New Hampshire event and ii) people seem to know and care a lot about AI, I’d reckon that he’d likely talk about alignment, and do a much better job of it than most others.

3. He seems to have pretty cool blog posts.

Is there any way I can remove the "community posts" section from my front page? I used to have the "community" tag blocked completely, so now I'm seeing a lot more community posts than I'd like.

It's in Cavendish! A long-term goal is to beat them in Nobel prizes..

The first time I heard about effective altruism was when someone told me "you should check out your local EA community--it's where the cool people hang out". Indeed this turned out to be the case; I had met thoughtful, curious people, and gotten quite interested in the ideas of effective altruism.

When I moved a year later to a different city in a different country, I spent a few weeks lamenting the difficulty of meeting people, until I realized I could go have thoughtful and interesting conversations with wonderful people at the local EA community. In a way, it feels not unlike to churches in towns in the United States--you can move from one state to another, and yet the next Sunday you'll find a tight-knit community of people who approximately share a large subset of your values. And while in some denominations of Christianity, there's a large global hierarchical structure, as far as I am aware, there's no global fund for starting new churches (I may be totally wrong, I don't actually know much about this. Though I have heard of church planting, but even there, the new church "must eventually have a separate life of its own and be able to function without its parent body").

One of the major reasons people I've met valued the global EA community was due to the coordination problems of otherwise getting a lot of people to start some new initiative. That is, say you're trying to start an organization centred around giving RL agents numinous experiences. The median number of people who think this is the most worthwhile thing to work on, in each local EA community, is exactly zero, but through the connections formed by the global community (through EAG and forums) it's a lot easier to find people interested in starting rLANE, which would never have been possible on the local scale.

Fundamentally, at my level of involvement with the EA community, I'm not sure I've witnessed the failure modes that have been taken as the antecedent for this whole post. Small, local communities haven't felt elitist in the same way that nobody calls the local swing dance club elitist, despite it being very obvious whether you're an experienced swing dancer or not. The global community has been very good at solving coordination problems to tackle specific challenges, and I haven't witnessed any downsides personally (but once again, I've never been accepted to an EAG or anything, and I've never lived in any of the "big" EA cities). If decentralization is the correct direction to go down, I think the "who's interested in helping me work on <x>?" is one of the most valuable things to make sure has a strong foundation of interconnectedness between different communities in different places.