Cross-boost (excerpts) from
v buckenham (@v21@.com)
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.
...
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...
A similar thing for #coding is already getting problematic, apparently:
(new GPT learning from incorrect samples generated by a previous GPT and posted all over the web)