Eleos
Research Scientist
Berkeley, Remote. Posted 5 months ago
The role
You will conduct foundational and applied machine learning research into the potential well-being and moral status of AI systems. This work happens at the intersection of machine learning, cognitive science, and philosophy of mind to inform decision-makers and policymakers.
What you will do
- Design and conduct empirical research on foundational topics like LLM preferences, values, introspection, persona vectors, and valence representation
- Design and conduct applied research on the potential well-being of frontier AI systems
- Collaborate with external researchers across the AI welfare and AI safety ecosystems
- Write papers and blog posts
- Present talks reporting research
Required qualifications
- Strong interest in understanding the potential well-being of AI systems
- A track record of leading and publishing impactful empirical ML research
- The ability to design, run, and iterate on experiments quickly and creatively
- Engineering skills to implement necessary tooling and infrastructure
Nice to have
- Familiarity with existing research on AI welfare and/or safety
- A PhD in computer science or the cognitive sciences
- Significant experience with mechanistic interpretability methods
Benefits
- Comprehensive health, dental, and vision insurance
- Employee premiums covered 100% by Eleos
- 75% of dependent premiums covered
- 401(k) with immediate vesting and 100% employer match up to 6% of salary
- Unlimited paid time off
- Parental leave for new parents
- Catered meals
- Snacks and drinks
- Gym and shower
- Relocation support stipend for moving to the SF Bay Area
- Health and wellness benefits
Before you apply
- Open to U.S. citizens, permanent residents, and individuals currently authorized to work in the U.S.
- Visa sponsorship may be available depending on the situation
About the company
Eleos is a nonprofit organization working to understand and address the potential well-being of AI systems. They take a pragmatic approach to research regarding the moral status of near-future AI systems.