Project summary

Rachel is a technical AI safety researcher and PhD student at the Center for Human-Compatible AI (CHAI) at UC Berkeley. She researches problems with reward learning as a path to AGI alignment and mentors promising junior researchers through CHAI internships and one-off advising for 80,000 Hours.

This grant will cover her medical and productivity expenses so that she can work full time on technical AI safety research & mentorship. She has had health problems since 2020, and because of her low stipend and inflexible choice of insurance, has been struggling to pay for the necessary medical care. This has already reduced her productivity and the time she can devote to her highest-impact work, and will continue to do so if these expenses aren’t covered.

Project goals

Without additional funding, Rachel would have to take up additional work in EA operations or tutoring, or more likely, switch to a full-time internship in tech or finance for a few months, pulling her away from her much higher-impact AI safety work. This grant is meant to help her devote more time to her highest-impact work.

Additionally, getting upfront funding may allow her to undergo treatments and recover more quickly, which means she can work more total. Finally, this will reduce her financial stress, which could otherwise cause burn out or at the very least distraction, also diverting her from her work.

How will this funding be used?

In order of priority, so the top of the list is covered first, and the lower items get covered as more money is received:

• $15,000: medical expenses

• $10,000: buffer for miscellaneous and unanticipated medical costs

• $9,000: timesaving services like a PA (who would do primarily medical-related admin work) & experiments with promising health interventions like sleep study or physical therapy

• $9,600: taxes for this grant

• additional funding: meal delivery services, personal cleaning, and other time-saving services.

In her Nonlinear application, Rachel requests $43k ($34k post-tax), and estimates that she could productively use up to $58k ($45k post-tax).

How could this project be actively harmful?

Don’t see how it could be.

What other funding is this person or project getting? Where else did this person or project apply for funding in the past?

Rachel received a Nonlinear Support Fund grant, which reimburses up to $10k for coaching, tutoring, and other services that do not overlap with the ones being funded by this grant.