Stay Up to Date
Join newsletter subscribers.
Subscribe to our RSS feed.Subscribe
Follow us on social networks.
If progress in artificial intelligence continues into the future, AI systems will eventually surpass humans in intelligence. Most leading AI researchers expect this to happen sometime this century. MIRI’s goal is to get a head start on the technical obstacles to making smarter-than-human artificial agents safe and robust.
We focus on three basic categories of open problem: designing autonomous agents that highly reliably pursue some goal; value specification for such agents; and making such agents error-tolerant.
For an in-depth introduction to some of the problems we are working on in these three areas, see our research agenda. We also have a research guide for people considering getting started in the field.
There are ample opportunities to collaborate with us on our ongoing projects. If you are interested in decision theory, reasoning under uncertainty, or other areas discussed in our research guide, you’re welcome to get in touch with us through our application form.
The quickest ways to get involved are:
Workshops: We host a number of multi-day math research workshops each year in the US and Europe. Past workshops have resulted in papers about probabilistic logic, program equilibrium, andobstacles to consistent self-reflection. We’ll notify accepted workshop attendees when relevant workshops are scheduled to take place, which can be many months down the line.
Forum: The online Agent Foundations Research Forum hosts active discussions and unpolished, preliminary results related to long-term AI alignment research. If you’re a new visitor, you’ll be able to start discussions by submitting external links that you have made.
MIRIx: From casual dinner conversations about MIRI papers to multi-day research workshops, MIRIx events are independent MIRI-funded discussion groups. Anyone can become a MIRIx event organizer, and in some cases we’ll be able to send a MIRI research fellow to a meeting to give tutorials and answer questions.
Research Assistance: Our research assistance program encourages students in mathematics, computer science, and formal philosophy to collaborate and consult with our researchers to help steer and inform their work on AI alignment problems.
If you are interested in any of the above programs or opportunities, please apply below so we can chat with you about your specific interests and availability.Apply now to participate in MIRI’s research
If you are a researcher in computer science, logic, or mathematics with a proven track record of high-quality research, we’d love to connect with you about the career opportunities at MIRI. Note that Research Fellowship applicants are typically required to first attend a Research Workshop.
We are seeking researchers who can collaborate to solve open problems in aligning AI with human values, and prepare those results for publication. Researchers will be selected for their background, ability to generate productive new research ideas, and proactive attitude toward research.
The ideal candidate has a career interest in these research questions and aims to develop into a senior research fellow at MIRI, or aims to continue these avenues of research at another institution after completing substantial work at MIRI.
Through the end of August 2015, MIRI is running a fundraiser to help us scale up our research team. Interest in long-term AI safety engineering is growing quickly, and MIRI is in an excellent position to direct this momentum toward high-priority technical work. How quickly we are able to expand will depend on how many of our funding targets we are able to hit: