Summary: CLR is hiring for our Summer Research Fellowship. Join us for eight weeks to work on s-risk motivated empirical AI safety research. Apply here by Tuesday 15th April 23:59 PT.

We, the Center on Long-Term Risk, are looking for Summer Research Fellows to explore strategies for reducing suffering in the long-term future (s-risks) and work on technical AI safety ideas related to that. For eight weeks, fellows will be part of our team while working on their own research project. During this time, you will be in regular contact with our researchers and other fellows, and receive guidance from an experienced mentor.

You will work on challenging research questions relevant to reducing suffering. You will be integrated and collaborate with our team of intellectually curious, hard-working, and caring people, all of whom share a profound drive to make the biggest difference they can.

While this iteration retains the basic structure of previous rounds, there are several key differences:

• We are particularly interested in applicants who wish to engage in s-risk relevant empirical AI safety work.^[1]
• We encourage applications from individuals who may be less familiar with CLR’s work on s-risk reduction but are nonetheless interested in empirical AI safety research. To facilitate this, we have shortened the first round of the application process.
• We are especially looking for individuals seriously considering transitioning into s-risk research, whether to assess their fit or explore potential employment at CLR.
• We expect to make significantly fewer offers than in previous rounds, likely between two and four, with some possibility of making none. This is due to limited mentorship capacity.

Apply here by Tuesday 15th April 23:59 PT.

We're also preparing to hire for permanent research positions soon. If you'd like to stay informed, sign up for our mailing list on our website. We also encourage those interested in permanent positions to apply for the Summer Research Fellowship. 

Further details on the fellowship here

1. ^{^}

   We are currently undergoing a strategic shift in our research priorities. Moving forward, the majority of our work will focus on s-risk-motivated empirical AI safety research in the following areas:

   • Personas/characters – How do models develop different personas or preferences? What are the most plausible training stories by which models develop malevolent or otherwise undesirable personalities? What preferences will misaligned models have by default, and what affordances will developers have to influence those preferences even if alignment does not succeed?
   • Multi-agent dynamics – How do models behave in extended multi-agent interactions, especially adversarial interactions where agents have conflicting goals? How well can we predict the behaviour models in extended multi-agent interactions from their behaviour on shorter and cheaper evals (e.g., single-turn evals)?
   • AI for strategy research – How can (future) AI assistants meaningfully contribute to macrostrategy research or other forms of non-empirical research? How could we verify that AI assistants were producing high-quality macrostrategy research?

   For more details on our theory of change and our general approach to empirical s-risk research, please see our measurement agenda (although note our focus has somewhat narrowed since publishing that).

   We will also continue to explore s-risk macrostrategy, with a particular focus on understanding when and how interventions in AI development can robustly reduce s-risk. While we may accept some summer research fellows to work on this area, we expect most fellows to focus on the empirical research agenda.