AI Safety Camp

AI Safety Camp (AISC) is a non-profit initiative that runs programs for diversely skilled researchers who want to collaborate on open problems for reducing AI existential risk. The program was conceived in November 2017 at Effective Altruism Global London by Linda Linsefors, who was seeking a collaborative research environment for AI safety work. Together with Remmelt Ellen, who took on meetings and logistics, the first camp was organized and held in April 2018 in Las Palmas, Gran Canaria, with 21 participants working across 5 research teams over 10 days. Since then, AISC has run 11 editions. The early camps (AISC1-AISC3) were held in person in locations including Las Palmas, Prague, and Madrid. Starting with AISC4 in 2020, the program shifted to a fully virtual format, which it has maintained since. Each edition runs for approximately 3 months, typically from mid-January to mid-April, with all participants committing at least 10 hours per week and attending weekly meetings. Project leads propose and oversee research initiatives, while team members join projects based on their interests and skills. The program has produced significant outcomes across its history. Alumni have gone on to take 43 jobs in AI safety at organizations including OpenAI, Anthropic, DeepMind, MIRI, and various governance bodies. Ten organizations have been founded by alumni, including Arb Research, AI Safety Support, and AI Standards Lab. Projects originating at AISC have received a total of $1.4 million in follow-on grants, and seven published papers have emerged from camp projects. An independent impact assessment by Arb Research found that 5-10% of participants become AI safety researchers who would not otherwise have entered the field, at a cost of approximately $12,000-$30,000 per new researcher. The organizing team is led by Remmelt Ellen, who has been with AISC from the start and oversees program design and strategy, and Robert Kralisch, who focuses on onboarding diverse projects with an emphasis on conceptual progress on the AI Alignment Problem. Linda Linsefors, the co-founder, has transitioned to an advisory role. Additional organizers contribute part-time to applications, team selection, participant facilitation, and outreach. AISC has historically faced funding challenges despite strong community endorsement of its impact. The program has received major grants from the Future Fund ($290,000), Effective Altruism Funds ($180,000), and the Survival and Flourishing Fund ($130,000) as of July 2022. The program nearly shut down in early 2024 due to funding shortfalls before successful community fundraising efforts kept it going. AISC operates with a variable budget ranging from $15,000 for a minimal edition to $300,000 for a fully stipended program.