News for hackers. Stuff that matters.

A comment on Neural Networks Are Already Clean. Parent comment

Re:Neural Networks Are Already Clean by lukewarm707 on Friday October 16, 2026 @06:52AM (#50116803)
"a network’s own mechanisms are its [...] building blocks" is tautological.

> > "the standard evidence we had for that belief was misunderstood."

The author did NOT misunderstand the evidence of polysemanticity and superposition. Rather, the explanatory hypothesis was incorrect.

[ Parent | on HN ]
Re:Neural Networks Are Already Clean by mike_hearn on Friday October 16, 2026 @07:53AM (#50117269)
The word "legible" is the critical part of that sentence.
[ Parent | on HN ]
Re:Neural Networks Are Already Clean by lukewarm707 on Friday October 16, 2026 @09:07AM (#50117923)
No: "a network’s own mechanisms are its legible building blocks"

Yes: "mechanisms within networks are legible"

or even, "networks are more legible than we thought".

[ Parent | on HN ]