A community for upskilling in AI safety research engineering
“… To continue making this progress, we need more engineers.”
- Anthropic; “The engineering challenges of scaling interpretability”
Research engineering skills are among the main bottlenecks in AI safety - leading organisations such as Anthropic and the AI Safety Institute are constrained by how many individuals can rapidly iterate on experiments and scale them up to larger models.
If you are a STEM researcher, software engineer or data scientist looking at a career in AI safety, then building up research engineering skills is an excellent next step.
The Research Engineers Club is a group for individuals who want to pursue this path.
The idea behind the Research Engineers Club is to partner up individuals with complementary skills, and facilitate knowledge transfer through paper replications by pair-programming.
To launch the Research Engineers Club, the aim is to begin developing a library of “Research Engineering Guides” based on papers recommended by our expert advisors; these guides will be produced by the founding cohort for the benefit of future members of the Club.
The founding cohort will also have the opportunity to:
Attend the keynote address by Dan Braun, Lead Engineer at Apollo Research
Use a dedicated channel in the SAIL Discord to discuss projects with others in the founding cohort
Attend in-person engineering sprints at the London Initiative for Safe AI (LISA) coworking space
Be reimbursed for compute costs and hours spent developing the Research Engineering Guides
Engage the AI safety research community to devise extensions which build on the results of their chosen paper
Applications to join the founding cohort are now closed, you can express interest in joining future cohorts below.