Why are we not using AI for this?
This is the sort of thing that AI excels at.
Link: https://www.politico.com/news/2026/01/22/epstein-files-release-review-00739987
Replies to: Why are we not using AI for this?
It's too busy make memes of Trump-Beetlejuice roaming around Greenland.
(no message)
Depends. For classified files, you'd need a closed A.I. source.
(I realize you have expertise in this area and probably already know this...)
But obviously they wouldn't just scan them into a public domain source like ChatGPT to get a summary.
Correct, you train a model on this data specifically.
It's really the only way that anyone will be able to make sense of 2M documents. There entirety of Wikipedia is only like 7M articles.
I also don't expect that DOJ lawyers are going to be able to redact all of the appropriate names (victims) with any degree of accuracy. But LLMs are made for tasks such as this.
there have to be a lot of variables there that could cause serious liability issues if mishandled.
it could involve judgement and/or adjudication (ie human intervention) at which ai might not be proficient enough yet.
Humans have already messed this up.
(no message)
so who trains the model? it would be a totally new model, n'est-ce pas?
(no message)
Gov IT with sufficient clearance.
Anyone can with technical skills can train and host their own model privately, there are a number of open source tools for doing this.
Will AI excerpt out Bill Gates’ name as a professional courtesy?
(no message)
That would be considered a feature, not a bug.
(no message)
Consent Management