OpenAI
Researcher, Alignment
$295K–$500K
San Francisco · Full-time · Posted Aug 27, 2024
- Education
- PhD or equivalent experience in research in computer science, computational science, data science, cognitive science, or similar fields.
About the job
Conduct alignment research by designing, implementing, and evaluating experiments that help AI systems follow human intent and remain safe as their capabilities grow.
What you'll do
- Develop alignment capabilities, evaluations, and tools to measure risks, human-intent alignment, and model robustness.
- Design experiments on how alignment scales with compute, data, context and action lengths, and adversary resources.
- Develop human-AI interaction and scalable oversight methods, train models to assess correctness and risk, and explore AI applications in alignment research.
What they ask for
- PhD or equivalent experience in research in computer science, computational science, data science, cognitive science, or a similar field.
- Strong engineering skills, particularly designing and optimizing large-scale machine-learning systems.
- Deep understanding of the science behind alignment algorithms and techniques.
Nice to have
- Experience with PyTorch.
- Ability to develop data visualization or data collection interfaces using TypeScript or Python.
Pay
USD $295K – $500K base salary
Tools
PyTorch, TypeScript, Python
AI focus
Alignment research, including evaluation and intervention methods, scalable human oversight, model robustness, and using AI in alignment research.
Model-written summary · the original posting is authoritative