-
text generated by models trained on text generated by ... @mildlydiverting/1490297528922820611Permalink Look on archive.org
♻️ 39 Retweets ❤️ 225 Favorites Mood 0
-
there is a real fear among AI researchers that the last big corpuses of human written text have already been captured. all future scrapes of the internet for text to learn from will be contaminated by machine-speak.Permalink Look on archive.org
-
(let's leave aside the implicit horror in the way a stray Reddit comment, a Wikipedia edit, this tweet are captured and used to smooth or wrinkle a machine model that will be used to generate more text in the service of... who? whoever the researchers sell the model to)Permalink Look on archive.org
-
CAPTCHAs at least make it obvious that our unpaid micro-labour is being used in the service of capital @v21/944298512694554625Permalink Look on archive.org
-
anyway, funny to think of a time when generated text is recognizable due to it's use of typically 2020-ish speech patterns and references. a cultural fixed point new models start from, slowly becoming more evident as human culture drifts away.Permalink Look on archive.org
-
"context drift", as per this genuinely horrifying sci fi short story : qntm.org/mmacevedoPermalink Look on archive.org
-
of course, this kind of text model is not a flat and neutral capture of 2020, either. continuing in the grand tradition of psychology experiments, it represents, at an average, a white American college student. with a slight bias towards "the kind of person who goes on Reddit"Permalink Look on archive.org
-
i love this comparison @ethanplaut/1577616000845250560Look on archive.org