Research Engineer / Scientist, Alignment Science

Position Overview

• Location Type: Hybrid (Preference for Bay Area, open to 25% travel)
• Employment Type: Full-time
• Salary: Not specified

Anthropic's mission is to create reliable, interpretable, and steerable AI systems. We want AI to be safe and beneficial for our users and for society as a whole. Our team is a quickly growing group of committed researchers, engineers, policy experts, and business leaders working together to build beneficial AI systems.

This role involves building and running machine learning experiments to understand and steer the behavior of powerful AI systems, with a focus on AI safety and risks from advanced systems. The ideal candidate is both a scientist and an engineer, dedicated to making AI helpful, honest, and harmless.

Requirements

• Significant software, ML, or research engineering experience
• Some experience contributing to empirical AI research projects
• Familiarity with machine learning concepts and techniques

Responsibilities

• Build and run elegant and thorough machine learning experiments.
• Contribute to exploratory experimental research on AI safety.
• Collaborate with other teams, including Interpretability, Fine-Tuning, and the Frontier Red Team.
• Focus on risks from powerful future systems (like ASL-3 or ASL-4 under the Responsible Scaling Policy).
• Participate in projects such as:
  • Testing the robustness of safety techniques by training language models to subvert interventions.
  • Running multi-agent reinforcement learning experiments (e.g., AI Debate).
  • Building tooling to efficiently evaluate the effectiveness of novel LLM-generated jailbreaks.
  • Writing scripts and prompts to efficiently produce evaluation questions to test models’ reasoning abilities.
  • Contributing ideas, figures, and writing to research papers, blog posts, and talks.
  • Running experiments that feed into key AI safety efforts, like the design and implementation of the Responsible Scaling Policy.

Company Information

About Anthropic:

Anthropic’s mission is to create reliable, interpretable, and steerable AI systems. We want AI to be safe and beneficial for our users and for society as a whole. Our team is a quickly growing group of committed researchers, engineers, policy experts, and business leaders working together to build beneficial AI systems.

Current Topics of Focus:

• Scalable Oversight: Developing techniques to keep highly capable models helpful and honest, even as they surpass human-level intelligence.
• AI Control: Creating methods to ensure advanced AI systems remain safe and harmless in unfamiliar scenarios.
• Alignment Stress-testing: Creating model organisms of misalignment to improve our empirical understanding of alignment failures.
• Automated Alignment Research: Building and aligning a system that can speed up & improve alignment research.

Application Instructions

• Not specified in the provided text.