#AcademicRunPlaylist - 10/30/24

I found this cool machine in a local library, and while unfortunately it doesn't print out academic talks I did listen to a lot of them for my #AcademicRunPlaylist!

First was a nice talk by Rohith Kuditipudi on distortion-free mechanisms for language model provenance at the Simons Institute for the Theory of Computing https://www.youtube.com/watch?v=fYrvZx7VERI

Next was an interesting talk by William Schueller on analyzing open source software development at the Santa Fe Institute https://www.youtube.com/watch?v=LzA2J9Rs8RE

Next was an informative talk by Pamela Samuelson on the state of copyright lawsuits against generative AI companies at the Simons Institute https://www.youtube.com/watch?v=S7Zp_vGUnrY

Next was an engaging talk by Donald Johanson on the discovery of Lucy and paleo-anthropology more broadly at the ASU Institute of Human Origins https://www.youtube.com/watch?v=BjOg1BVyjhA

Next was an amazing talk by Noah Golowich on a novel approach for robust LLM watermarking at the Simons Institute. This method is extremely elegant and compelling, hopefully it's taken up by vendors quickly. Highly recommend https://www.youtube.com/watch?v=VFt7NYGj_F4

Next was a great talk by Hiroki Habuka and Tatsuhiko Inatani on Japan's approach to AI governance at the Stanford Cyber Policy Center https://www.youtube.com/watch?v=4O6OhZOhRyM

Next was an intriguing talk by Miranda Christ on pseudorandom error-correcting codes for generative AI watermarking (particularly images) at the Simons Institute https://www.youtube.com/watch?v=UMSIc5USW5Y

Next was a wide-ranging talk by Jesus Solis on how US military bases in east Asia (particularly Japan) spawned a huge black market system at the Weatherhead Center for International Affairs at Harvard University https://www.youtube.com/watch?v=J1sgI2sh7js

Next was a short talk by Scott Aaronson on the next steps for generative AI watermarking at the Simons Institute https://www.youtube.com/watch?v=YzuVet3YkkA

Next was a fantastic talk by Cynthia Rudin on why simpler machine learning models often perform better than black box ones (and why we shouldn't necessarily expect the reverse) at UW Statistics. There's also a mini-concert at the end with songs about machine learning. 'nuff said. Highly recommend https://www.youtube.com/watch?v=qAFDv5JfhPc

Last was an excellent talk by Frauke Kreuter on the challenges of developing systems that are universally "aligned" at the Simons Institute https://www.youtube.com/watch?v=4FrP3QMLWUw