Disclaimer: all of these are judgements are based on a particular context (Cambridge), though I think they generalise across local AIS Groups. I make assumptions about how EA is perceived in AIS pollicy circles that may be miscalibrated. Tell me if I'm wrong.
EA Cambridge has a funny problem. Many of the most engaged, most value aligned committee have decided they should work on AIS, and are interested in helping run Cambridge AIS Hub.
Obviously I should be excited! Committee have done their diligent cause prio, and come to a defensible solution. I should talk to them to help them figure out if this is the most impactful move. I should raise my reasonable concerns about EA Cambridge losing it's most high context members. We should come to some conclusions that respect their autonomy and their belief in the goodness of EA and EA Cambridge's longevity.
However, I should also be worried! Is this what I should be doing? Should any EA be doing principles first EA? Should we all do AIS?
Before you stop me to ask, "but what about the arguments for the importance of EA in the face of AGI". Id like this post to avoid retreading well worn ground. I'm not going to address all of EA's place in the AI Age.
Instead I'm going to posit a few things that make me believe some people in AIS should proudly wear the EA badge because this reinforces what makes the best AIS groups to well. That their members think about hard about how they can do the most good. This thinking might begin to happen when we remind AIS organisers that these motivations spur the AIS scene to distinguish itself from AI ethics or other less scope sensitive movements, otherwise concerned about AI.
I think many more should use EA principles, and discuss AIS work in EA terms to each other, even if they don't declare EA identification publicly.
I sometimes find it hard to believe that, in Cambridge, an EA group is a full time organiser’s best use of time, if that full time organiser could plausibly work on AIS field building. Empirically, at Cambridge the AIS group has more demand, the people in its programs have more knowledge about AIS than the EA group has about EA. Have more agency or at least demonstrate more willingness to do things like write publicly, put their hands up in socials.
It seems like, if you want to get people into things that are good for the world, and you care about tractability, you'd do better at the AIS group. Almost all the things you would like a good group to be, you'd have more success with via focusing on AIS. If you're NEST(https://frommatter.substack.com/p/opening-a-soul-saving-shop), the most close knit, "take the problems" seriously, group of AIS people would be more close knit, more serious. I'd guess you'd get more Macaskill types
If you're CEA, you'd get more people into impactful careers at great organisations, more quickly. They'd just all be AIS organisations.
But wait! AIS organisations keep telling us that a great indicator for value alignment in a good hire is being an EA. Some hiring people even say EA aligned people are needed for various stakesy roles. Especially the leadership ones. What gives?
EA motivated thinking, in the style that the best AIS people do, is excellent for mitigating criticisms of AIS' as a field's effectiveness. When you consider theories of change, you do better work that's less likely to be harmful. When you career plan with an eye for neglectedness, you realise you shouldn't just jump on the technical AIS bandwagon because you're doing computer science.
I think there's been a sticky info hazard, that's worked it's way into lots of AIS minds. That for people to find out that you're an EA in AIS, is like your colleagues to find out you're a terrorist in the civil service. There are good cases where EA's reputation pushes away opportunities. Say in certain political arenas. But you should figure out what theories of change, your endorsement of EA has, negatively and positively for your impact. Those negative cases, where Alice called herself an EA in her blog, and this got her rejected from a private dinner with big policy people, are the vast minority of cases for a given person wanting to get into AIS. The positive cases meanwhile, as a person in a local group, may be absurdly strong. You validate prioritisation, and rigor, and actually making sure what you're doing is the best thing you could be doing. You bring in considerations of space faring factory farming. Of longtermism (which still motivates much of the AIS movement even though it rarely comes up in Introductory courses).
The reason I wrote this, is because of this post. The three AGI pills. It talks about the differences between people who believe in AI, AGI and ASI. People in CAISH seem to be on this spectrum without realising it. That is to say there is a missing mood. People often talk past each other about things because they don't have the language to confront. Important questions about what actually matters. Are they talking about ASI or merely AGI? How does this translate into the interventions that actually attack those risks? Instead, there is a focus on the newest safety card, on the latest METR report. These are flashy but they don't help people decide what parts of AIS they think actually matter most, given what they think the world will look like, and given their stances on longtermism, on S-risks Vs x-risks. This thinking might begin to happen when we remind AIS organisers that these motivations spur the AIS scene to distinguish itself from AI ethics or other less scope sensitive movements, otherwise concerned about AI.
I'm moving to AIS, partly because I think having proud EAs in AIS is how I can do the most good right now.