Research Scientist, Machine Learning
OpenAI
·
Posted:
May 19, 2023
·
Onsite
About the position
The Policy Research team at OpenAI is seeking an experienced machine learning researcher to shape and lead the ML research agenda for trustworthy AI. The role involves researching and developing novel evaluation methods and interventions for dangerous model capabilities, existential risks, fairness and representation, and undesired model behavior. The successful candidate will work with downstream product and infrastructure teams to build and scale effective tools for responsible deployment, and mentor ML Researchers on the Deployment Planning team. The role reports to the Trustworthy AI lead and is based in San Francisco.
Responsibilities
- Research upstream interventions at the level of training data, pre-training, and training
- Research and prototype novel evaluation methods in areas such as dangerous model capabilities and existential risks, fairness and representation, as well as untruthful, hallucinatory, or otherwise undesired model behavior.
- Work with downstream product and infrastructure teams to build and scale effective tools for responsible deployment
- Develop and mentor ML Researchers on the Deployment Planning team
- Architect and develop interventions that improve real world impact
Requirements
- Have a track record of coming up with new ideas or improving upon existing ideas in machine learning, demonstrated by accomplishments such as first author publications or projects
- Possess the ability to own and pursue a research agenda, including choosing impactful research problems and autonomously carrying out long-running projects
- Have experience developing novel techniques for ML model measurement and mitigation
- Have experience in research mentorship, leading project teams, and setting technical direction
- Be comfortable working cross functionally across both research and product teams
- Past experience in interdisciplinary research collaborations (nice to have)
- Past experience in creating high-performance implementations of deep learning algorithms (nice to have)