Women in AI Safety Hackathon

Shape the future of safe and ethical AI development! Whether you're a researcher, developer, policy enthusiast, or new to AI safety - join us for an empowering weekend of innovation and collaboration. No prior AI safety experience required. Together, we can build the foundational tools and frameworks needed for responsible AI development.

We're thrilled to partner with Women Who Do Data (W2D2), a member-led community dedicated to supporting diverse talent in AI technology development, to bring you this unique hackathon during International Women's Day weekend. Together, we're working to increase the presence of underrepresented groups in AI safety and technical AI development.

We're excited to announce that Lambda Labs will be providing $400 in computing credits to each participating team! This generous support will give you access to powerful cloud instances (including A100s) to help bring your ideas to life. Each track will have a $600 prize for the winning team, with a total of $1,800 in prizes across all three tracks!

The Women in AI Safety Hackathon brings together talented individuals to tackle crucial challenges in AI development and deployment. This event particularly encourages women and underrepresented groups to contribute their unique perspectives to critical areas of AI safety, including alignment, governance, security, and evaluation.

As AI systems become increasingly powerful and pervasive, diverse perspectives in their development and safety mechanisms are more crucial than ever. This hackathon provides a platform for participants to:

• Collaborate with leading women researchers and practitioners in AI safety
• Develop practical solutions to pressing AI safety challenges
• Build lasting connections in the AI safety community
• Receive mentorship from experienced professionals
• Present ideas to industry experts

Challenge Tracks

Mechanistic Interpretability Track

We're excited to announce that Goodfire, a pioneering research lab in AI interpretability, will be sponsoring our Mechanistic Interpretability track! We're also thrilled to have Myra Deng, Founding PM at Goodfire, join us for a HackTalk on understanding and steering AI models.

1. Understanding AI Model Internals: Dive deep into the inner workings of large language models using state-of-the-art sparse autoencoder techniques. Learn to map and understand model behavior at a granular level, while developing tools to interpret and visualize neural features. This track focuses on creating innovative approaches to understand what's happening inside these complex systems.

2. Model Steering and Editing: Explore practical applications of interpretability by developing methods to modify and control model behavior through targeted feature interventions. Work on creating interpretable control mechanisms and safety-focused editing tools that can help shape model outputs while maintaining performance. This component bridges the gap between theoretical understanding and practical applications.

3. Feature Analysis and Visualization: Create tools and interfaces that make AI systems more transparent and understandable. Focus on building interactive visualization dashboards that help analyze feature activations and interactions, making complex neural networks more accessible to researchers and practitioners. Develop novel ways to present and interact with model internals

Participants will have access to:‍

- Goodfire's Ember API for model interpretation
- Compute credits for experiments

The winning team in this track will receive a $600 prize!

Public Education Track

‍
We're excited to announce that BlueDot Impact, the world’s leading AI Safety education platform with a community of over 4,500 professionals, will sponsor our Public Education track!

Build an education platform feature to help newcomers to the field understand the risks of AI.

Ideas to get you started include:

1. Games: Get creative and build an exciting new feature to demonstrate or solidify understanding of core AI Safety concepts for non-technical learners
2. Matching platform: We are a global community with participants from over 100 different countries. Introduce a mentorship matching feature to connect mentors and mentees with similar interests from across our community
3. Self-assessments: How can our course graduates self-test understanding of critical AI Safety concepts and demonstrate mastery?
4. Journey mapping: It can be difficult for newcomers to see how they can fit in the larger world of AI Safety. Create a matching tool to navigate through AI safety domains, job recommendations, or BlueDot course offerings based on a user's background, skills, and interests
   

The winning team in this track will receive a $600 prize!

Social Sciences Track

This track invites participants to explore the intersection of Artificial Intelligence and the Social Sciences through two broad themes: (1) how AI "thinks" and (2) how it shapes human society.

The first theme, Machine Psychology, examines the systematic study of AI behaviour and cognition, drawing insights from psychology, behavioural science, and cognitive science. This theme investigates how AI models develop internal representations, exhibit emergent capabilities, display decision-making patterns, and adapt to different inputs. It explores both the interpretability of these systems and their behavioural tendencies, helping us understand how AI "thinks”.

The second theme, AI in Society, focuses on the human-AI dynamic, investigating AI’s broader social, political, and economic implications. This includes its impact on human interactions, institutional transformation across sectors like education and healthcare, and evolving social norms. The theme also addresses critical ethical considerations in AI development and deployment, examining questions of fairness, accountability, and governance as AI systems become more integrated into society.

Participants will contribute through short reports or research-based papers that can range from highly technical to completely non-technical, as we want to make this track accessible to all social scientists and foster interdisciplinary discussions on AI’s evolving role in our world.

The winning team in this track will receive a $600 prize!