Unable to load image

Daily Bard Digest 2024-08-18

Live feed of the best moderator on the internet, straight into your veins.

Here's todays official post from the BARDCHIVE:

06/21/15 21:32:22 with a score of -1: https://old.reddit.com/r/OutOfTheLoop/comments/3anadz/what_does_being_triggered_mean_in_all_those/csedt1x/?context=8

I meant exactly what I wrote.

Dictating to someone that they didn't mean what they wrote while misquoting them (I wrote "medically diagnosed individuals", not "medically diagnosed people"), even after they've asserted that no, they didn't mis-state, and downmarseying with your sockpuppet ("No, I mean exactly what I wrote." is currently at -1) is the height of stupid arrogance.

I wrote "… or related condition …" in my original comment. That is what I was referring to.

It's clear that you're a fatpeoplehate troll. I would ask you to kindly frick off but I'm not certain that you're old enough that this wouldn't technically be soliciting a minor, so I will suggest that you do your Geometry I homework instead of weakly trolling this thread.

1
Jump in the discussion.

No email address required.

At 08/18/24 09:50:26 in technology with a score of 1 point:

AI models have the extremism trained out of them

No. They have a set of front-end & back-end safeguards & filters that detect & redirect from extremist inputs & extremist outputs, but importantly, because these are human-written filters overlaid on the inputs & outputs, they are not capable of detecting all extremist prompts or directives, nor all extremist outputs, and can be circumvented by creative prompting & can be removed entirely by people hosting their own instances of the software. And the filters can only interdict overt, historically proven expressions.

They're still recommending putting glue on pizza. They've been trained on content that itself relies heavily on context, which context has been collapsed due to being distilled into a probabalistic model. And a huge quantity of what those models were trained on are unlabeled and undetected hate speech, harassment, violent threats, and antisocial tendencies

Jump in the discussion.

No email address required.

Link copied to clipboard
Action successful!
Error, please refresh the page and try again.