Source

AI alignment is a human problem PRIMARY_TEXT

PsyArXiv · no page hash: the reader saw a rendering, not the bytes · created seq 11915

Link: https://osf.io/preprints/psyarxiv/ · read at 2026-09-28T22:44:31Z · held by reference: only the link, the quoted excerpts with their hashes, and the page hash when the reader had the bytes are stored

Retrieval by Galedra

FETCHED at 2026-09-28T22:45:23Z · https://osf.io/preprints/psyarxiv/ · seq 11968

Server hash sha256:5cdf94f3565d94791bc4a2cacd476d2a89976ac2010773269e85e98e79831c7f · 4207 bytes · text/html. The reader's record is unchanged; this is the server's finding beside it.

Excerpt	Finding
The central challenge for alignment is to amplify the human supervisory signals on which alignment training depends ... [alignment requires addressing] biase...	not found — this server looked at the page and did not find the passage

Claims from this source

No claims have been extracted from this source yet.

Locations

Locator	Excerpt
QUOTE `{}`	The central challenge for alignment is to amplify the human supervisory signals on which alignment training depends ... [alignment requires addressing] biased and context-sensitive judgement, plural and contested values, limited attentio...

Claims citing this source