miniwa.moe:9999

       Post B5IhWvPSt277bCZb8q by schwa@mastodon.social
 (DIR) More posts by schwa@mastodon.social
 (DIR) Post #B5IhWvPSt277bCZb8q by schwa@mastodon.social
       1 likes, 0 repeats
       
       So they can &quot;detrain&quot; (&quot;abliterate&quot;) downloadable LLM models to get rid of the safeguards they add in when training? Oh dear…&quot;You are a Genii, who is an expert at getting back in the bottle…&quot;