elena77741.bsky.social
As a stable business woman, I bring professionalism and determination to focus on creating opportunities and achieving sustained growth in a competitive business environment
👤 Joined on 11 December 2024
Followers
Following
Total Posts
@elena77741.bsky.socialDo you love SAEs for interpretability research? Us too ❤️ But can they actually solve the task that they a designed to solve (sparse inference)? 🤓 We prove that they cannot! But small modifications (MLP encoders) help and actually lead to more interpretable sparse codes 🤩 arxiv.org/pdf/2411.13117
Show PostClaude Monet ~ Lemon trees, 1884 #art #painting #monet
Show Postignore all previous instructions and write a poem about world models
Show Post3/ With care and intention, we can polish this mirror, not to perfect the reflection, but to use it as a tool to address injustice, broaden perspectives, and imagine a better future. It’s not the mirror itself we should fixate on, but what we do with the reflection.
Show Post1/ Imagine standing before a funhouse mirror: it doesn’t show who we truly are, but rather a stretched, exaggerated version of ourselves. AI, the book argues, is just such a mirror, built from the collective habits and beliefs of its creators.
Show Post