Yeah yeah AI system has false positives. Like a human, really. In fact humans are pretty bad at detecting AI text: https://arxiv.org/html/2211.13087
Among AI models, Blenderbot stood out: in AI-AI conversations, it was judged human 67% of the times - more often than actual human-human conversations. In human-AI conversations, human participants were labeled as humans 68% of the time, and AIs were classified as AI 56% of the time.
(FYI, a detection rate of 50% is the same as random chance, so 56% is pretty poor)
First, we’re the humans. Investments into the existence and wellbeing of humans don’t have to be justified by an ROI. It’s kind of all we’ve been doing since we came out of the caves.
A tool on the other hand needs to justify it’s existence by saving us more time and money than it cost to make or enabling us to do things we couldn’t do before, otherwise it’s a shitty tool.
Second, the LLM that claimed Frankenstein was slop took tens or hundreds of billions to make. And that cost doesn’t scale with the demand. To be able to run even one instance of it you have to spend the entire lump sum of training the model. On the other hand you can train one more human if you need one more skilled worker and you don’t have to redevelop the entire curriculum from scratch to do that.
Yeah yeah AI system has false positives. Like a human, really. In fact humans are pretty bad at detecting AI text: https://arxiv.org/html/2211.13087
(FYI, a detection rate of 50% is the same as random chance, so 56% is pretty poor)
The point of the post is there is no way to reliably detect it as it just reproduces what it was trained on.
A human also did not take a trillion $ of investment to be able to produce this false output.
a human takes tens (maybe hundreds depending on school) of thousands of dollars of training and can’t even be copied
First, we’re the humans. Investments into the existence and wellbeing of humans don’t have to be justified by an ROI. It’s kind of all we’ve been doing since we came out of the caves. A tool on the other hand needs to justify it’s existence by saving us more time and money than it cost to make or enabling us to do things we couldn’t do before, otherwise it’s a shitty tool.
Second, the LLM that claimed Frankenstein was slop took tens or hundreds of billions to make. And that cost doesn’t scale with the demand. To be able to run even one instance of it you have to spend the entire lump sum of training the model. On the other hand you can train one more human if you need one more skilled worker and you don’t have to redevelop the entire curriculum from scratch to do that.
I hope people realize this before they call out other comments as “AI generated”, something I see all too often on Lemmy