Plans

AI Safety Camp serves a purpose within a wider ecosystem. Aspirants go through stages of our program to test their fit for jobs at the frontiers (for research and also indirectly for operations, entrepreneurship, and funding grants).   

To run each edition, we rely on the support of…

  • research mentors (experienced alumni, senior researchers)  to select and mentor participants around topics they prioritise.
  • ops assistants (local-community builders, charity support staff)  to assist participants and reliably handle event logistics.
  • advisors (fellow entrepreneurs, research staff managers)  to coach us on improving our program and strategy.
  • funders (‘earn-to-give’ donors, grantmakers)  to cover the costs of each edition.

Are you interested to eg. mentor a team, host a venue, partner on a program, or make a tax-deductible donation? Email contact@aisafety.camp

 

Looking back

Diverse organizers ran 5 editions since our pilot in 2018, mostly funded through grants from the Long-Term Future Fund/CEA.
Private donors funded remaining budget gaps – to ensure we could cover the venue, catering, and individual reimbursements.

 

Looking forward

We are professionalising. Our focus for 2021 is to more consistently target and serve…

  • participants who have the capacity and need to try collaborating on AI existential safety research alongside
  • mentors who have built expertise around a research agenda they are excited to guide such a team on.

We ran edition #5 in the first half of 2021, which went smoothly. In 2022, we will run a virtual camp (#6) and a California coast camp (#7).

 

Ongoing considerations

We are resolving a few remaining concerns (see 2021 updates below) with our advisors and the LTFF around whether…

  • mentors can provide adequate guidance in ways they find a good use of their time.
  • participants think through priorities and work out research projects to be useful for the community’s efforts.
  • the camp experience benefits those participants who go on to do particularly promising research. 
  • we’ll have enough funding to start paying lead volunteers to dedicate their time and attention. 

Where we decide to go from here, in consultation with our partners, depends amongst others on the extent to which…

  • mentors become more involved in selecting research problems they are particularly excited about and willing to guide teams to work on.
    updates:  AISC5 proposals got more feedback but three teams missed a mentor. AISC6 participants instead select a mentor’s problem.
  • next participants get a better grasp of AI-alignment concepts they’re unfamiliar with (added this point later)
    update:  AISC6 participants get some pre-reading and can then deliberate ways to apply concepts & considerations at the first weekend.
  • next participants are more satisfied during a camp and afterwards more able to contribute to research (see internal metrics)
    update:  Some alumni surveyed by LTFF reported very substantial positive benefits from attending a camp. Others reported no benefit.
  • past leads submit their interest to manage areas of responsibility they excelled at as volunteers.
    update:  Three new contractors lead AISC6 and may become permanent coordinators depending on progress on their set responsibilities.
  • LTFF decides to ramp up grants
    updates:  LTFF granted $85K (via AISS);  SAF granted $35K (via SEE);  SFF granted $130K (via RCAISS)

We may hire up to four permanent coordinators to oversee research mentoring, participant facilitation, event logistics, and the overall program.