News for hackers. Stuff that matters.

A comment on Show HN: Llama 3.2 Interpretability with Sparse Autoencoders. Parent comment

Re:Show HN: Llama 3.2 Interpretability with Sparse Autoencoders by fsndz on Thursday November 21, 2024 @07:55PM (#42210215)
I stopped at: "causal sequence of “thoughts” "
[ Parent | on HN ]
Re:Show HN: Llama 3.2 Interpretability with Sparse Autoencoders by benchmarkist on Thursday November 21, 2024 @08:23PM (#42210349)
Interpretability research is basically a projection of the original function implemented by the neural network onto a sub-space of "explanatory" functions that people consider to be more understandable. You're right that the words they use to sell the research is completely nonsensical because the abstract process has nothing to do with anything causal.
[ Parent | on HN ]
Re:Show HN: Llama 3.2 Interpretability with Sparse Autoencoders by HeatrayEnjoyer on Thursday November 21, 2024 @10:00PM (#42210754)
All code is causal.
[ Parent | on HN ]
Re:Show HN: Llama 3.2 Interpretability with Sparse Autoencoders by benchmarkist on Thursday November 21, 2024 @10:05PM (#42210772)
Which makes it entirely irrelevant as a descriptive term.
[ Parent | on HN ]
Re:Show HN: Llama 3.2 Interpretability with Sparse Autoencoders by mdp2021 on Friday November 22, 2024 @04:26AM (#42212381)
"Servers shall be strict in formulation and flexible in interpretation."
[ Parent | on HN ]