On this #InternationalArchivesWeek I'd like to repost this absolute gem by the @bucksarchives.bsky.social!
🤬
You do not "fill the gap" in archives.
You ask *why* the gap is there in the first place.
That is where the real information resides.
Also:
LLMs train on digitized materials.
Do you have any idea how little has been digitized?
LLMs train on a *sliver* of the human past and human knowledge.
For #InternationalArchivesWeek, consider how digitized copies of manuscripts can preserve access to a manuscript's contents, provide a record of ownership if the object is removed, and provide access to cultural heritage if a community is displaced.
Learn more: hmml.org/stories/reversal-of-fates
Rebranding a hallucination as a ‘critical confabulation’ that can fill archival gaps is a big no from me.
The more things change, the more they stay the same...
Nice study that essentially corroborates Michel Rolph Trouillot's silencing framework.
The diagram embodies all silencing processes (fact creation/assembly/retrieval/retrospective significance), a bleak reminder on how power operates today.
Amalia S. Levi, PhD
Amalia S. Levi, PhD
Peter Whitewood
Amalia S. Levi, PhD
HMML
Hoping this helps our colleagues across the industry
1/ Excited to report we have a new paper out
@nature.com today! The bottom line: training data for LLMs does not just fall from the sky - it is created in the context of existing social political institutions - and that has consequences for LLM output.
nature.com/articles/s41...
nature.com
Government-controlled media influences the output of large language models via their training data, and models queried in the languages of countries with lower media freedom show a stronger ...
