<html><head><meta http-equiv="Content-Type" content="text/html; charset=UTF-8"></head><body dir="auto"><div dir="auto">For images OpenAI has added watermarks to identify AI generated images. But these watermarks are apparently easy to remove.</div><div dir="auto">https://gizmodo.com/openai-chatgpt-how-to-check-watermark-metadata-c2pa-1851234655</div><div dir="auto"><br></div><div dir="auto">For texts it is harder to recognize if they are AI generated or not unless it is a regurgitation of a learned text. The lawsuit of the NY Times against OpenAI is based on 100 examples where AI generated text matches NY Times articles verbatim word for word. </div><div dir="auto">https://www.vox.com/technology/2024/1/18/24041598/openai-new-york-times-copyright-lawsuit-napster-google-sony</div><div dir="auto"><br></div><div dir="auto">-J.</div><div dir="auto"><br></div><div><br></div><div align="left" dir="auto" style="font-size:100%;color:#000000"><div>-------- Original message --------</div><div>From: glen <gepropella@gmail.com> </div><div>Date: 2/8/24 5:54 PM (GMT+01:00) </div><div>To: friam@redfish.com </div><div>Subject: [FRIAM] provenance </div><div><br></div></div>She Was Falsely Accused of Cheating With AI — And She Won’t Be the Last<br>https://www.rollingstone.com/culture/culture-features/student-accused-ai-cheating-turnitin-1234747351/<br><br>That is an old article. But what interests me is: "Then she began gathering evidence that she’d written the brief herself."<br><br>It strikes me that the best way to give evidence that you actually wrote something (or generated any kind of [cough] content), is to show its evolution from start to finish. For things like code and text, it seems obvious to simply use a revision control system. That can include something as irritating as Track Changes in Word, Git, or whatever. But for blobs, it seems more difficult. Most tools like GIMP have an Undo function, but only for that session. If you close the project then re-open it, you can't undo what you did before you closed it. It seems reasonable to store some sort of trace, even if it's prohibitive to allow for full reversibility.<br><br>A second-best way might be to keep either all your old content so you can extract patterns from it and, then, show how some one instance of it matches your induced "signature". Or you could throw away or assume old content will be lost and keep a defeasible inference database, continually trained up on your content, handy to give such pattern-oriented evidence. But this sort of diachronic database has the same flaws bemoaned in the article (e.g. TurnItIn's AI detector). Plus, some of us want to doff such histories once and awhile, make ourselves anew.<br><br>But if we were to encourage people of, say, age 10 to store their works, we'd have to teach them how to index their portfolio. We'd have to teach library science at a very early age. I can't even find notes I took a week ago, much less crap I wrote decades ago.<br><br>-- <br>ꙮ Mɥǝu ǝlǝdɥɐuʇs ɟᴉƃɥʇ' ʇɥǝ ƃɹɐss snɟɟǝɹs˙ ꙮ<br><br>-. --- - / ...- .- .-.. .. -.. / -- --- .-. ... . / -.-. --- -.. .<br>FRIAM Applied Complexity Group listserv<br>Fridays 9a-12p Friday St. Johns Cafe / Thursdays 9a-12p Zoom https://bit.ly/virtualfriam<br>to (un)subscribe http://redfish.com/mailman/listinfo/friam_redfish.com<br>FRIAM-COMIC http://friam-comic.blogspot.com/<br>archives: 5/2017 thru present https://redfish.com/pipermail/friam_redfish.com/<br> 1/2003 thru 6/2021 http://friam.383.s1.nabble.com/<br></body></html>