What subjects should I study in university to prepare for alignment research?

Fields that are likely to give you relevant background understanding to work in alignment include mathematics, computer science, or artificial intelligence itself. However, AI alignment

also benefits from having researchers with diverse backgrounds, so if you have a particular interest or talent in a different field, it can be valuable to pursue a degree in that field instead. Alignment has had contributions from the fields of neuroscience, philosophy, biology, cybersecurity, risk management, safety engineering, and economics, and it has been argued in particular that AI safety needs social scientists.

Technical content that is broadly useful to learn in university courses includes:

• Machine learning
  Machine learning

  An approach to AI in which, instead of designing an algorithm directly, we have the system search through possible algorithms based on how well they do on some training data.

  , especially a firm grasp of the basics of deep learning and transformers in particular, as well as reinforcement learning
  Reinforcement learning

  A machine learning method in which the machine gets rewards based on its actions, and is adjusted to be more likely to take actions that lead to high reward.

  View full definition
• Mathematics, including statistics, linear algebra, and calculus

Other subjects that are relevant to alignment include:

• Topics in economics, such as game theory, resource allocation, incentives, and cooperation;
• Topics in philosophy, such as semiotics, agency, ethics and rational rigor;
• Topics in biology, such as neuroscience and evolution;
• Engineering fields, such as information theory

In addition to taking relevant courses, it is also helpful to study AI safety

materials outside of university.An excellent place to start is the AI safety fundamentals course. This post on the EA forum gives some more broad advice about pursuing alignment as a career.

Some words of more general advice: It can be easy to get stuck in a strategy of upskilling that becomes disconnected from the final goal of working on AI alignment. It is often helpful to keep your final goals in mind, and “backchain” to figure out what you need to learn and do.

If your final goal is working on alignment research, this might look like learning about alignment specifically and using that to figure out what you want to study. It might also look like keeping this final goal in mind when approaching different courses. You can also get in touch with others who are interested in AI safety, such as by joining a local group, a community related to AI safety – or as a quick default, the AI Alignment Slack.

A degree and lots of formal education are not necessary to do high quality research, and many of the biggest contributions have been made by people without a traditional academic background. There may be contributions you can make today!