Concrete, clear, and concise wins the day. We expect to get a lot of applications; you’ll stand out if you can crisply describe your background, what you plan to work on concretely, and why it matters for AGI alignment. Keep your main application short; you can link to or attach longer additional materials.
Most of all: great technical research!
Otherwise: We ask for a brief email every quarter updating on your progress during the term of your grant. And we expect you to publish or otherwise make publicly available your research results at the end of your grant.
We think smart graduate students can often make large contributions. So we want to empower them directly!
The $150k OpenAI Superalignment Fellowship covers a $75k stipend and $75k in compute and research funding. We’ll work with you on your individual circumstances, but we generally hope to pay out the compute funding to graduate students directly, rather than going through your university; we want to graduate students themselves to be able direct what the compute is used for, rather than other intermediaries.
This fellowship is for one year in the first instance, but if this goes well, we hope to continue the program. If you are a graduate student but have differing needs from this default fellowship (e.g., larger compute needs) you can still apply for a normal grant as well.
We’re really excited to see what you do!
We’ll invite grant winners to a Slack with other grant winners, OpenAI researchers, and other researchers working on alignment. We hope that this will be a place to exchange research ideas and intermediate updates. We also plan to arrange office hours and other opportunities to discuss research with OpenAI researchers.
Yes, that’s fine. (But do note the other submissions in the form, just so we have an overview.)