Palisade Research Research Associate Berkeley, CA · Full time Company website

Apply for Research Associate

Palisade's mission is to help people and institutions build the understanding needed to avoid permanent disempowerment by strategic AI agents. We research the capabilities, motivations, and drives of frontier AI systems. Our work has included demonstrations of specification gaming, shutdown resistance, and autonomous hacking and self-replication. In addition to the research literature, our work has been cited in congressional hearings, the Wall Street Journal, MIT Technology Review, the BBC, and elsewhere.

We are hiring an early-career researcher to help us update and extend our existing research results.

We have a collection of research results we’ve produced in the past, and we need your help to bring them up to date with new developments. We want to focus on finding and correcting methodological weaknesses, robustness issues, or outdated assumptions, as well as updating and expanding that research to include newer models, better techniques, and related work from the broader community.

We’ll want to publish clear, high-quality updates (on arXiv, our blog, and LessWrong), and this will require fluency with academic English and comfort writing for technical audiences.

You will work with the Head of Research and other members of the research team - the role would benefit most from someone with a track record of relevant research, who is ready to take substantial ownership of projects like this.

Responsibilities:

• Examine and extend experiments to look at newer models; explore variant prompts, improved scaffolding, stronger baselines; related methodological refinements
• Review past research for errors, quality issues, and opportunities for improvement, using frontier LLMs to assist with analysis, critique, and re-testing
• Engage with external work that builds on and critiques Palisade’s findings; incorporate the useful ideas into updated evaluations.
• Write and publish clear updates, corrections, and extensions (preprints, LessWrong posts)
• Help maintain high standards of research integrity and transparency (reproducibility, clear documentation of methods and limitations, etc.)

Desired Qualifications:

• Some research background in a technical field.
• Strong command of academic English and the ability to produce clear writeups, suitable for arXiv, technical blog posts, and LessWrong.
• Basic experience with computer programming and the technical fundamentals of modern AI systems.
• Experience running LLM experiments (prompting, evaluation, simple scaffolding or agent setups, analyzing results).
• Careful, skeptical approach to empirical claims.
• Interest in the strategic capabilities, motivations, and controllability of advanced AI systems.

Details of Position:

• Full-time, either hybrid or in-person in Berkeley, CA. Strong remote candidates will be considered.
• Reports to the Head of Research, Ben Weinstein-Raun
• All job-related inquiries should be directed to [email protected] to ensure fair consideration. We're unable to discuss openings through other channels.
• Salary: $100k-200k, depending on location and experience
• Benefits: Comprehensive health, dental, and long-term disability coverage; unlimited PTO
• Equal opportunity employer.

Palisade Research is an equal opportunity employer. We do not discriminate on the basis of race, color, religion, sex, gender identity or expression, sexual orientation, national origin, ancestry, age, marital status, medical condition, physical or mental disability, genetic information, veteran status, or any other characteristic protected by federal, state, or local law.

Pursuant to the California Fair Chance Act and the Berkeley Fair Chance Access to Employment Ordinance, we will consider qualified applicants with arrest and conviction records.

If you need a reasonable accommodation during the application or interview process, contact us at [email protected].