Krishnan Srinivasan

About

Krishnan Srinivasan is a researcher working at the intersection of robotics, reinforcement learning, and artificial intelligence, with particular expertise in safe RL, dexterous manipulation, and multimodal learning. He is perhaps best known as a co-author of the landmark 2021 report "On the Opportunities and Risks of Foundation Models," an enormously influential work that coined the term "foundation models" to describe large-scale, broadly trained AI systems like GPT-3 and DALL-E — a paper that has accumulated over 2,100 citations and helped reshape how researchers conceptualize modern AI development. Beyond this seminal contribution, Srinivasan has made meaningful advances in safe reinforcement learning through Recovery RL, an algorithm that enables agents to explore uncertain environments while respecting safety constraints — earning nearly 200 citations. His work on contact-rich robotic manipulation explores how robots can integrate vision and touch through self-supervised multimodal learning, addressing longstanding challenges in dexterous in-hand control. He has also contributed to assistive robotics, developing learned latent action spaces that help users with disabilities teleoperate complex robotic arms more intuitively. Collectively, his research reflects a commitment to making intelligent robotic systems both capable and trustworthy in real-world settings.

Research Focus

Key Achievements

Top Papers

1. 1
   On the Opportunities and Risks of Foundation Models

   2,177 citations · 2021
2. 2
   Recovery RL: Safe Reinforcement Learning With Learned Recovery Zones

   193 citations · 2021
3. 3
   Learning Hierarchical Control for Robust In-Hand Manipulation

   34 citations · 2020
4. 4
   Making Sense of Vision and Touch: Self-Supervised Learning of Multimodal Representations for Contact-Rich Tasks

   32 citations · 2019
5. 5
   Recovery RL: Safe Reinforcement Learning with Learned Recovery Zones

   20 citations · 2020
6. 6
   Using a Variable-Friction Robot Hand to Determine Proprioceptive Features for Object Classification During Within-Hand-Manipulation

   15 citations · 2019
7. 7
   Making Sense of Vision and Touch: Learning Multimodal Representations for Contact-Rich Tasks

   14 citations · 2020
8. 8
   Swarm Robotics: A New Framework of Military Robots

   12 citations · 2021
9. 9
   Controlling Assistive Robots with Learned Latent Actions

   8 citations · 2020
10. 10
    Dexterous Manipulation Primitives for the Real Robot Challenge

    4 citations · 2021

Key Collaborators

Contact & Links