#AcademicRunPlaylist - 9/23/24

It was a nice fall day in Boston, and I enjoyed the first leaves of the fall with some talks for my #AcademicRunPlaylist!

First was a pair of talks by Charlotte Bradley (re-theatricalizing AI) and Sayak Roy (inhabiting AI through routine improvisation and disordering) at the Design Justice Network AI Institute https://www.youtube.com/watch?v=VO49TyHWJL0

Next was an excellent talk by Boris Hanin on neural network scaling limits at Harvard CMSA. Hanin builds up an intuition for the relationship between data, network depth, and network width to understand the limits of very large models, an essential approach for designing these models and evaluating claims about them. This is mostly a repeat of the talk he gave at Rutgers a few months ago, and if you didn't watch that one you owe it to yourself to catch this one https://www.youtube.com/watch?v=0PJJ29jYIsg&t=4s

Next was a great talk by Nikolai Matni on designing robotic systems with robust learning, control, and stability methods at the Carnegie Mellon University Robotics Institute https://www.youtube.com/watch?v=3z4XlWHF1II

Next was an interesting talk by Melanie Weber on data and model geometry in deep learning at Harvard CMSA https://www.youtube.com/watch?v=uwVEVSS8DnI

Next was a fascinating talk by Jerome Lewis on the long duration and resilience of hunter-gatherer egalitarian civilizations in Central Africa at UCL Anthropology https://www.youtube.com/watch?v=7qf30hOocV8

Next was a thought-provoking talk by Neil Thompson on the improvements in algorithm performance over the decades and implications for AI at Harvard CMSA https://www.youtube.com/watch?v=PfEYKyfu0O0&t=1s

Next was an intriguing talk by Anastasia Bizyaeva on nonlinear dynamics of beliefs over social networks at the Women in Network Science Seminar https://www.youtube.com/watch?v=sPDpAA7M_NM

Next was a fantastic talk by Jamie Morgenstern on what leads to disparity in modern machine learning applications at Harvard CMSA. Starting from basic models, Morgenstern shows the different factors that lead to disparity reductions, extending this intuition to modern ML settings and the issues that presents. Highly recommend https://www.youtube.com/watch?v=GXnCPXeTb_s

Next was a wide-ranging talk by Vikash Mansinghka on embodied cognition at MIT Brain and Cognitive Sciences. This focuses a lot on probabilistic inference, as well as some practicalities for getting these models up and running yourself https://www.youtube.com/watch?v=jbiAONqWmbA

Last was a nice talk by Kavita Ramanan on understanding high-dimensional stochastic dynamics on epidemiological network models at Harvard CMSA https://www.youtube.com/watch?v=YIjd6ypI-pk&t=2s