Posted HW0 now and will post future ones also!
1Video is on youtube too
Thanks - dropped it for now
Posted our homework zero on the CS 2881 website, along with video and slides of last lecture and pre-reading for tomorrow's lecture. (Homework zero was required to submit to apply to the course.)
Actually I did want to link to the debate between Ajeya and Narayanan. As part of working on the website I wanted to try out various AI tools since I haven't been coding outside the OpenAI codebase for a while, and I might have gone overboard :)
Btw I assigned AI 2027 as pre reading for the first lecture! (As well as “AI as a normal technology “ and the METR paper on doubling time for tasks.)
First lecture for CS 2881 is online https://boazbk.github.io/mltheoryseminar/ + prereading for next lecture.
Students will be posting blog posts here with lecture summaries as well.
Thank you - although these type of questions (how can a weak agent verify the actions of a strong agent) are closely related to my background in computational complexity (e.g., interactive proofs, probabilistically checkable proofs, delegation of computation), I plan to keep the course very empirical
Yes there is a general question I want to talk about which is the gap between training, evaluation, and deployment, and the reasons why models might be:
1. Able to tell in which of these environments they are in
2. Act differently based on that
A blog post on the first lecture of CS 2881r, as well as another blog on the student experiment by @Valerio Pepe , are now posted . See the CS2881r tag for all of these
https://www.lesswrong.com/w/cs-2881r
The video for the second lecture is also now posted on YouTube