Overview
Location: Berkeley/San Francisco (may consider a remote option)
Duration: 10-14 weeks, most likely starting mid-June
Compensation: Full-time pay (minimum $35/hr), plus cost of travel. May include assistance with housing
Application deadline: March 15, 2022
About AI Impacts
AI Impacts conducts original research to answer decision-relevant questions about the future of artificial intelligence, and maintains an online knowledge base on these topics. Our projects involve a variety of disciplines, including artificial intelligence, neuroscience, history, economics, computer science, physical science, evolutionary biology and philosophy.
Carrying out this kind of work involves:
- reasoning about how to make open-ended questions tractable,
- evaluating the existing academic literature,
- gathering and analyzing new data of various kinds,
- writing in academic and encyclopedic styles.
Examples of work we have done or may do:
- Measuring computing performance trends, as input to forecasting future trends
- Reasoning about what human brains imply about the computing hardware that might be needed for ‘human-level’ AI performance
- Analysis of arguments for AI posing an existential risk to humanity
- Collection and analysis of empirical data on the frequency of discontinuous progress across a range of technologies
- Interviews on the plausibility of AI being relatively safe by default
- Survey of machine learning researchers on opinion about the future of AI
More of our past work can be found elsewhere on our website.
We aim to inform decision-making around artificial intelligence, by both others familiar with the topic (e.g. people in the AI safety community making choices about what scenarios to focus their efforts on), and people new to it (e.g. policymakers with broad mandates).
Our workplace:
- Is flexible and accommodating to idiosyncratic needs and preferences
- Prioritizes employee thriving
- Values openness, kindness, and truth-seeking: we strive to be a place where everyone can talk about real opinions and raise stupid questions
- Values epistemic carefulness and reasoning transparency, and also making rough and fast estimates where appropriate
The program
This is a summer internship for students from a variety of backgrounds who are interested in doing research related to threats from advanced artificial intelligence. Interns will work on their own projects, and in support of our ongoing work.
We will match interns to projects based on their interests and the tractability of the project. Examples of possible projects include:
- Writing code to scrape, analyze, and present data on AI performance over time
- Investigating and writing about how people approached problems in the past
- Estimating the rate at which various actors could scale up inputs to AI, such as semiconductor manufacturing
- Find data to test hypotheses about the mechanisms behind technological progress
The program may also include:
- Talks by AI risk professionals from outside AI Impacts
- An end-of-summer mini-conference where interns can present their work
- Single- or multi-day trips to destinations in the Bay Area
Selection criteria
This program is geared toward university students at any stage of undergraduate or graduate work and recent graduates, but we are open to anyone who is interested in working with us for a couple months over the summer. Importantly, you do not need to have any specific skills or background knowledge. Past researchers at AI Impacts have included people with backgrounds in journalism, physics, philosophy, and economics. What is most important is that you are interested in our work, are comfortable attacking open-ended problems outside your area of expertise, and can take a careful, systematic approach to thinking through these problems.
You must be authorized to work in the United States. If you are on a student visa, we recommend checking with your university.
How to apply
If you are interested, please fill out our application by 15th of March. We may keep applications open later than this, but we do not guarantee considering applications past this date. We will likely reach out to a first round of strong applicants by early March. If you have any questions or are uncertain about how to fill out the form or whether you may be a good fit, please email Rick at rick@aiimpacts.org.