3 ms·

Your understanding of society’s use of AI is not in line with reality. Your opinion that a 23% true positive rate with a 9% false positive rate is ok is not in

Your understanding of society’s use of AI is not in line with reality. Your opinion that a 23% true positive rate with a 9% false positive rate is ok is not in line with general principles (not even just Western-centric) of the burden of proof of guilt. Only 23% of US adults have tried ChatGPT,[1] so to say that we live “in a world where so much AI in human writing” as you do in another comment is simply false. Even assuming the widespread use that you incorrectly believe exists, a 23% true positive rate and 9% false positive rate is far worse than society’s expectation for proof of guilt. >It is better that ten guilty persons escape than that one innocent suffer.[2] Take a school class where no students used AI to cheat. Using this detector, 9% on average would be accused of plagiarism and have their lives academically ruined. That is not acceptable. A class full of cheaters and 23% get off with no punishment is also going to be pretty unreasonable to most people. [1] https://www.pewresearch.org/short-reads/2024/03/26/americans-use-of-chatgpt-is-ticking-up-but-few-trust-its-election-information/ https://www.pewresearch.org/short-reads/2024/03/26/americans... [2] https://en.m.wikipedia.org/wiki/Blackstone%27s_ratio https://en.m.wikipedia.org/wiki/Blackstone%27s_ratio

benreesman 3y ago
You can just have different standards for when you apply it: I want this thing on political ads. Don’t fire people or accuse them of plagiarism because of it. That would be stupid no matter how good it was.
ben_w 3y ago
In fairness, > in a world where so much AI in human writing Is not a percentage, and also the point (I think) isn't "how many people are using it" but "how much content has each produced", which is only close to equal when a human uses it to automate the output they would have created by themselves anyway. I do not know how many words have been written by LLMs vs. humans in the last year; as I have almost nothing to ground an estimate with, I can easily believe that humans are 3 orders of magnitude greater or lesser in output — one extreme bound due to the low price of tokens, the other extreme bound due to the high price and limited supply of hardware.

benreesman 3y ago

You can just have different standards for when you apply it: I want this thing on political ads. Don’t fire people or accuse them of plagiarism because of it. That would be stupid no matter how good it was.

ben_w 3y ago

In fairness, > in a world where so much AI in human writing Is not a percentage, and also the point (I think) isn't "how many people are using it" but "how much content has each produced", which is only close to equal when a human uses it to automate the output they would have created by themselves anyway. I do not know how many words have been written by LLMs vs. humans in the last year; as I have almost nothing to ground an estimate with, I can easily believe that humans are 3 orders of magnitude greater or lesser in output — one extreme bound due to the low price of tokens, the other extreme bound due to the high price and limited supply of hardware.