tl;dr: ACS Research is hiring for several full-time research positions in macrostrategy, LLM psychology, and ML. 1-2 year contracts in Prague, London, or San Francisco Bay Area, $80-300k. This could be a Pareto-optimal option for some of you based on research topics, incentives, salary, collaborators, and research environment. Details and applicartions here - deadline approaching soon.

Also please do refer people to us: we don't offer referral bounties, but hope you can tell people to apply just because it is good idea. In the rest of the post is a bit more about the roles, why you may want to apply and why not.

A bit longer form about the roles

Gradual Disempowerment Research Fellow We're looking for polymaths who can reason about civilizational dynamics. This role comes with a lot of intellectual freedom - it could mean economic modelling, theoretical work on multi-agent dynamics, historical analysis, and more.

LLM Psychology & Sociology Researcher We want people with a strong intuitive understanding of LLMs to help run empirical studies on topics like LLM introspection and self-conception, LLM social dynamics, and how ideologies spread between AIs.

AI Psychology & Agent Foundations ML Researcher We need people who can bring technical and methodological rigour, taking high-level ideas about AI psychology and turning them into concrete ML experiments. This could include of evaluations, mech interp, post-training, both APIs and open-weight models.

Why work at ACS and why not

Reasons why may include

• what we work on are some of the best questions in terms of importance, tractability and neglectedness
• we have better ability to forecast stuff than typical superforecasters ^[1]
• understanding the SOTA and sometimes future technology on a technical level  ^[2]
• network of collaborators and research partners
• quite a lot of flexibility
• freedom to express views publicly
• fairly good incentives (we mostly care about understanding things, and to some extent figuring out how to actually improve them)
• we do not focus on that but we do publish academically legible papers, so you have more traditional academic jobs as a backup

Rough picture of what to expect is a small team of polymaths, quite a lot of travel, and span of thinking from "what's wrong with prevailing theories of agency" to "why this peculiarity in this model output" and "what should we recommend to this regulator"

Why not

• it's certainly not a 9-5 job, with a nice office, manager who tells you what to do, etc
• lots of travel
• the topics we work on are a mixture of some of the most intellectually exciting questions ever and staring into the abyss directly - this is sometimes psychologically hard
• some of the topics we work on don't have great feedback loops

Also: AMA in comments today or tomorrow.

1. ^{^}

   We don't spent much time producing legible forecasts for external audience, as this is not our comparative advantage. Still, there is some legible evidence ACS members and close collaborators are better than typical superforecasters.

2. ^{^}

   We broadly understood e.g. scratchpads, chain-of-thought, tool use, RL + LLMs as agency-increasing or importance of LLM personas for user engagement before this become public knowledge. We generally don't publish ideas which seem almost purely capabilities speedups.