Events, programs, and opportunities from MAIA. Join mailing list
Resources
Why AI safety matters, how to learn it at MIT, and how to keep up with the field.
Why care about AI safety?
MAIA's core mission is to empower MIT students to recognize and reduce the potentially existential risks posed by the development of powerful AI systems.
More capable AI systems could change how people work, conduct research, and make decisions. We study how to prevent serious harm from these systems, including:
- Loss of control: As AI systems become more autonomous and self-improving, while being deployed with super-human speed at super-human scales across all levels of society, maintaining meaningful human oversight becomes harder, and the consequences of failure become catastrophic.
- Geopolitical instability: Accelerating capabilities create pressure for rushed deployment and risky preemptive actions, raising the likelihood of accidents and conflict.
- Concentration of power: Even in optimistic scenarios where alignment is solved, transformative AI risks centralizing power in ways that are incompatible with meaningful notions of democracy and personal self-determination.
- Gradual disempowerment: People could gradually lose influence over important decisions as institutions rely more heavily on AI.
MAIA helps students explore these questions through AI Safety Fundamentals, research projects, and conversations with researchers. You can learn alongside other students and find a concrete problem to work on.
Start here
Start with our eight-week reading and discussion program. No prior AI safety background required.
Short curated list of articles and videos making the case for caution.
Video explanations of core alignment problems. Start with "Intro to AI Safety".
Why this century could be the one where transformative AI changes everything, and why that is not a comfortable thought.
A concrete, month-by-month scenario for how the next few years of AI could go.
The long-form problem profile: the arguments, the counterarguments, and what you can do about it.
The careful, step-by-step version of the argument, with probabilities attached to each step.
Learn
Curricula, classes, labs, and fellowships.
AISF Curriculum
AI Safety Fundamentals reading curriculum for fellows.
MIT & Harvard Classes
MIT and Harvard classes relevant to AI safety.
MIT Faculty and Labs
MIT faculty and labs working on AI safety and related areas.
AI Safety Fellowships
Fellowships and programs for getting involved in AI safety.
Research by MAIA Members
Papers published by MAIA members.
Organizations We Work With
Labs, nonprofits, and policy groups MAIA works with.
MAIA Merch
Explore our shirts, stickers, pens, and V0 AI safety playing cards.
Stay informed
Newsletters, podcasts, forums, and job boards for keeping up with the field.
Newsletters
- Transformer · Shakeel Hashim
- AI Safety Newsletter · Center for AI Safety
- Import AI · Jack Clark
- Don't Worry About the Vase · Zvi Mowshowitz
- 80,000 Hours Newsletter · 80,000 Hours
Podcasts
- 80,000 Hours Podcast · Rob Wiblin and Luisa Rodriguez
- Dwarkesh Podcast · Dwarkesh Patel
- AXRP · Daniel Filan
- The Cognitive Revolution · Nathan Labenz
Forums and maps
- Alignment Forum · Technical alignment research and discussion
- LessWrong · The broader community forum
- AISafety.com map · Every org, program, and resource in the field
Careers
- AI Safety Careers · Curated job board
- 80,000 Hours Job Board · Roles across labs, policy, and nonprofits
- MAIA Fellowships page · Our guide to research fellowships