I've found the following abstract frame/set of heuristics useful for thinking about how we can try to affect (or predict) the long-term future:  “How do we want to spend our precision/reach points? And can we spend them more wisely?” [Meta: This is a rough, abstract, and pretty rambly note with assorted links; I’m just trying to pull some stuff out and synthesize it in a way I can more easily reference later (hoping to train habits along these lines). I don't think the ideas here are novel, and honestly I'm not sure who'd find this useful/interesting. (I might also keep editing it as I go.] ---------------------------------------- An underlying POV here is that (a) scope and (b) precision are in tension. (Alts: (a) "ambition / breadth / reach / ...” — vs —  (b) “predictability / fidelity / robustness / ...”). You can aim at something specific and nearby [high precision, limited reach] or at something larger and farther away, fuzzier [low precision, broad reach]. And if you care about the kind of effect you’re having (you want to make X happen, not just looking for influence ~for influence’s sake), this matters a bunch. Importantly, I think there are “architectural” features of the world/reality[1] that can ease this tension somewhat if they're used properly; if you channel your effort through them, you can transmit an intervention without it dissipating (or getting warped) as much as it otherwise would. Any channels like this will still be leaky (and they’re limited), but this sort of “structure” seems like the main thing to look for if you’re hoping to think about or improve the long-term future.  (See a related sketch diagram here. I also often picture something like: “what levers could reach across a paradigm shift?” (or: what features are invariant in relevant ways?)) ---------------------------------------- Some examples / thinking this through a bit: 1. Trying to organize or steer a social movement (/big group of people) might extend your reach, but

Some women on the Facebook support group "Cluster Headache Patients" comparing labor pain to cluster headache pain: * "Honestly, I had a natural childbirth and a cesarean and cluster headaches are 10 times worse than both." * "2 unmedicated births for me. Would rather do that every day than have another cluster" * "every day though, really?" * "yes. I'd rather go through childbirth without pain relief than CH." * "tenfold worse than popping a baby out" * "Nah, labour/giving birth is a walk in the park compared to ch […] I was in labour with my son for nearly 3 days, then the midwife had to break my cervix with her hands, but I'd still rather do that again than have another CH" * "Labor pain doesn’t even come close to CH! I’d choose labor pain ANY day over suffering from another CH" * "CH is a million times worse" * "I had 4 children, 3 were natural. CH is worse." * "I'd rather have a baby. And my placenta tore during all natural childbirth." * "I gave birth to 4 different babies. The smallest being 8lbs 14oz. The biggest being 10lbs 15oz. I would much rather give natural birth all over again than a CH." * "I've had three babies—one was overdue and born with his arm over his head. Having a baby is still cake compared to clusters."   These are just a few. They go on and on. (So far only one woman claiming childbirth was worse, who "nearly bled out in childbirth, got an episiotomy with zero freezing/drugs.")

I was reading the AI-enabled coups report and one of the mitigations is roughly "build the AI model such that it refuses to do coups". If you can pull that off then it solves the problem, but it means the model is incorrigible, and therefore you have to exactly specify the correct values up front because you're locked-in. There might be some other report saying the way to avoid bad value lock-in is to make AI corrigible, and you can object to that by pointing out how it enables coups. You can't look at the problems separately, you have to consider them at the same time and find a solution that works for every problem.

The electric grid is a powerful, current AI safety policy opportunity. ~25% of the cost of data centres is electricity and the electric grid is heavily regulated across government agencies. In fact, the very reason why many people claim nuclear energy is over regulated is exactly why we might be able to regulate AI strongly via electric grid regulation. This means "the grid" could be a strong contender for a space to make rapid and robust progress on AI safety policy. As an expert in the electric grid with a strong interest in AI safety and resilience, I see an electricity sector full of opportunities for quick wins and actually moving the needle. While "high-level" policy interventions such as SB 53 are important, a drawback of these types of policy interventions is twofold: 1 - They attract enormous public scrutiny. Electric regulation, on the other hand, hardly make it into local newspapers or industry press even. 2 - They don't really move the needle. They are more aspirational and rely on outcomes in court cases, enforcement, etc. The electric grid already has "kill switches" installed and integration with national security. On the other hand, in the electric grid, one can plausibly pass very strong regulation with physical, actual AI safety levers. Some examples: A - Large consumers of electricity are critical to the grid. It is not unforeseeable at all that gov't bodies could require a "kill switch" for data centres, not because of AI safety but due to grid health.  B - The military is extremely focused on electrical grids. They are also a likely gov't body that would act quickly and decisively on perceived threats on the grid. They likely influence requirements for cyber security of the electrical grid. This can include monitoring of critical load (yes, data centres) and access to the above "kill switch".  These are just two of many ideas I have around making rapid, robust progress on AI safety via much less public scrutiny, and using existing pathways and

Just noticed that I tend to up/downvote and agree/disagree vote more or less depending on what the current vote count is at. Standard herding bias at work. Hoping that saying it out loud will make it weaker, and maybe other people can relate.