People here don’t seem to understand what LLM detection is. All it does is search for patterns that are very common in chatbot generated speech. It’s not some magical property that’s metaphysical in nature. Either the speech is written by a chatbot, or Carney naturally talks in this sort of vapid and content free fashion which is common for politicians to do.
The real tell with AI writing is in the substance. It’s the weirdly balanced, almost bloodless neutrality on complex topics, the total lack of any authentic personal stake or lived experience, and a distinct feeling that you’re reading a brilliantly comprehensive Wikipedia summary instead of a thought that formed in a human mind with memories, biases, and a body.
It’s obviously pretty reliable at statistically identifying patterns common to LLM generated text. Wikipedia, having had a problem with a flood of LLM written articles, has put put a whole detailed guideline of what these patterns are, and why they’re associated with LLM generated text. I implore you to spend at least a modicum of time to actually understand the subject you’re attempting to debate here.
At this point, I have no idea what you’re even trying to say here is. When you say stuff like ‘it doesn’t make it more reliable’, what do you mean by that?
If you agree that you can reliably detect LLM speech patterns, then do you agree or disagree that the speech contains many patterns that closely resemble LLM generated text?
AI LLM detection is notoriously unreliable.
you can try it yourself by putting text written by actual people and see how that goes https://app.gptzero.me/
That doesn’t make LLM detection any more reliable
People here don’t seem to understand what LLM detection is. All it does is search for patterns that are very common in chatbot generated speech. It’s not some magical property that’s metaphysical in nature. Either the speech is written by a chatbot, or Carney naturally talks in this sort of vapid and content free fashion which is common for politicians to do.
The real tell with AI writing is in the substance. It’s the weirdly balanced, almost bloodless neutrality on complex topics, the total lack of any authentic personal stake or lived experience, and a distinct feeling that you’re reading a brilliantly comprehensive Wikipedia summary instead of a thought that formed in a human mind with memories, biases, and a body.
I get what it is. Trying myself vs you trying it doesn’t make it more reliable.
It’s obviously pretty reliable at statistically identifying patterns common to LLM generated text. Wikipedia, having had a problem with a flood of LLM written articles, has put put a whole detailed guideline of what these patterns are, and why they’re associated with LLM generated text. I implore you to spend at least a modicum of time to actually understand the subject you’re attempting to debate here.
https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing
I know how LLMs work. Nothing you say is going to convince me that me trying it myself is going to be more reliable than you trying it.
Like, what are you even disagreeing with me on?
At this point, I have no idea what you’re even trying to say here is. When you say stuff like ‘it doesn’t make it more reliable’, what do you mean by that?
If you agree that you can reliably detect LLM speech patterns, then do you agree or disagree that the speech contains many patterns that closely resemble LLM generated text?
Really?
You try it -> it has a certain level of reliability.
I try it -> that reliability doesn’t change.
That’s the only point I’m making. You just love to argue.