🧩 Philosophy 3h ago · Jacob_Hilton

Mechanistic estimation for wide random MLPs

Less Wrong
View Channel →
Mechanistic estimation for wide random MLPs
Source ↗ 👁 0 💬 0
This post covers joint work with Wilson Wu, George Robinson, Mike Winer, Victor Lecomte and Paul Christiano. Thanks to Geoffrey Irving and Jess Riedel for comments on the post.
In ARC's latest paper, we study the following problem: given a randomly initialized multilayer perceptron (MLP), produce an estimate for the expected output of the model under Gaussian input. The usual approach to this problem is to sample many possible inputs, run them all through the model, and take the average. Instead

Comments (0)

Sign in to join the discussion

More Like This

Over Eight Months of Progress in Two: Analyzing the Mythos Preview Capability Jump
LessWrong · 4h ago
AI #167: The Prior Restraint Era Begins
LessWrong · 6h ago
How to get better at chess (and everything else)
LessWrong · 9h ago
Multipolar Civilisation Depends on Maintaining an Attacker’s Dilemma
LessWrong · 9h ago
📰
18th Century German Aesthetics
Stanford Encyclopedia of Philosophy · 12h ago
📰
Normative Economics and Economic Justice
Stanford Encyclopedia of Philosophy · 18h ago