I think it’s a feedback loop. AI is trained off publicly available datasets like House of Commons records so popular words only get more popular the more AI slop is in there, since LLMs fundamentally just predict the next word given the context without much “logic” behind it.
Given enough time this will make LLMs basically unusable as public data gets contaminated with AI slop. But unfortunately that will also mean the public data itself is basically unusable.
I think it’s a feedback loop. AI is trained off publicly available datasets like House of Commons records so popular words only get more popular the more AI slop is in there, since LLMs fundamentally just predict the next word given the context without much “logic” behind it.
Given enough time this will make LLMs basically unusable as public data gets contaminated with AI slop. But unfortunately that will also mean the public data itself is basically unusable.