First was an interesting talk by Pulkit Agrawal on generalizing low-level skills for #robots at the #UPenn #GRASPLab. For robots, step by step instructions aren't enough for them to generalize, but using simulation and teacher-student learning Agrawal demonstrates promising results. Highly recommend https://www.youtube.com/watch?v=JE1FgzF9ixY (2/10) #robotics

Next was a fantastic talk by Daniel Hashimoto on building teams for surgical translation of #AI at the #GRASPLab. This is a fascinating tour of the history of surgical technology, #data collection, and modern tools. Beyond that, the new approaches being applied to surgical understanding are compelling, and represent an excellent fusion of technology, management, and #medicine. Highly recommend https://www.youtube.com/watch?v=yvuxkof8tRY (6/10) #surgery #bioethics

Next was an interesting talk by Guy Hacohen on learning the dynamics of neural networks at the #GRASPLab. The idea here is to identify which training examples are important at different learning stages, with implications for training regimes in resource constrained environments and #data collection https://www.youtube.com/watch?v=lBwtFiwcWiQ (4/10) #AI