⁂ L. Rhodes"Slop" is a pretty effective dismissal at the moment, but I fully expect the term to get hijacked by marketing the same way "hallucination" did. Soon you'll hear CEOs boasting that their AI generated 62% less slop, or whatever.
It's all slop. The prettiest, most convincing genAI image in the world is still slop. A heart-breaking work of staggering genius produced by an LLM is still slop. Slop is slop not because it's poorly done, but because it's produced by a process that by definition cannot care about what it's creating.
No matter what shorthand we come up with to express our dissatisfaction with the premise of their technology, the move will always be to respond as though it referred only to the shortcomings of an given implementation, because the point is not to address our actual concerns but to relativize our legitimate greivances.
The more direct attack would be to take the shine off of their terminology, loading them with the negative associations we have for the technology. And being really precise about it. Saying "he writes like an LLM" not because they used em-dashes or stock phrases, but because it's clear they didn't care about the thing they were writing. Or that an image "looks AI generated" not because it's surreal or anatomically incorrect but because it's soulless and derivative. Spur people to think about values that matter and why you can't get them via prompt.
Curio@lrhodes
To the people pushing LLMs, “writing like an LLM” is something they have always considered normal.
Then you get people realizing that it being LLMs all the way down in those circles you could have just sent over that original 10 line summary directly:
https://tech.lgbt/@Natasha_Jay/114773255690468383
That really speaks to what your saying.
Natasha :mastodon: 🇪🇺 (@[email protected])
Attached: 1 image They are starting to get it ...
JP