The Algorithmic Alignment Group is led by Dylan Hadfield-Menell, an assistant professor on the faculty of Artificial Intelligence and Decision-Making in the Computer Science and Artificial Intelligence Laboratory (CSAIL) at MIT. His research focuses on the problem of agent alignment: the challenge of identifying behaviours that are consistent with the goals of another actor or group of actors. The Algorithmic Alignment Group works to identify algorithmic solutions to alignment problems that arise from groups of AI systems, principal-agent pairs (i.e., human-robot teams), and societal oversight of ML systems.
The postdoc will work with an interdisciplinary team from several institutions including DeepMind and the Cooperative AI Foundation, as well as with MIT faculty, staff, and students.
A core component of human intelligence is the ability to effectively cooperate across a broad group of agents in a variety of one or few-shot interactions where agents are interdependent. A key open problem in AI is the development of agents or populations that mimic this cognitive ability. This has the promise to improve design strategies for helpful and prosocial artificial agents and can shed light on the development of these abilities in humans. We are looking for a postdoc to lead innovative research on these problems, specifically through the lens of multi-agent reinforcement learning. Specific goals of the research include:
- The design and launch of a contest on multi-agent reinforcement learning approaches to Cooperative AI at a top-tier ML conference, to standardise approaches and infrastructure for cooperation research,
- Leading novel research on the incentives for cooperative behaviours,
- Leading on a contest summary paper, along with the participants and senior advisors, to distill what was learned about the relative strengths and weaknesses of different approaches to cooperative AI.
This postdoc position is highly technical and visible and will play an integral role in designing the contest. The postdoc will work with an interdisciplinary team from several institutions including DeepMind and the Cooperative AI Foundation, as well as with MIT faculty, staff, and students.
The successful candidate will be expected to start in either September 2022 or January 2023. This is a temporary position with funding initially provided for 12 months, with a funding renewal process due to take place 9 months after the start date.