[kevindamm]

Where's the boundary, though? If someone generated slop but edits every sentence replacing at least half the words and ensuring it is in their voice consistently, does it still need the badge? If only one word is replaced but it corrected the hallucination and is otherwise reviewed for approval? If dice are rolled and the x'th word of the y'th sentence chosen for replacement?

I don't justify starting with slop in my own writings but I don't know whether you could even reliably label it appropriately. Even more so, it would be a shame to see genuinely human writing mischaracterized as genAI, especially in a public forum like LinkedIn.

[a3w]

At work, we had video content with "ai generated" now shown in an eponymous Teams channel. Now no one viewing the video knows if just the image, or the project logo, and/or the voiceover was AI generated but is very confused on that upon seeing the label.

Skip the labels. Photoshop and "the trainee did it" existed for 38 years already, and respectively for many more years now, and have about the same reliability.

[elpakal]

That's telling me the label was over applied, not that the label itself was not important. I'd love something that just tells me the likelihood that a text post was AI generated to start (like on LinkedIn, for eg).

[kevindamm]

Agreed. I think we'll see a further strengthening of reputation as a signal over any isolated statement or marketing.

[elpakal]

Then make it configurable per user, so people that dont want it can turn it off. Regarding the boundary idk but I still would rather know the likelihood that the content was AI generated (especially if it's high) than not.