I’m with Paris on the AI identification. This is just a tag that says whether the text, image or other media was AI generated. I think this is a good thing, it is certainly something I want to know going in.
If, as @JeffJarvis says, somebody who is bad at writing gives @paris a text that has been written or co-written by AI, she needs to know that. It doesn’t mean she should reject it out of hand, but it means she has to go back and ask the source additional questions, to ensure the veracity of the text.
Paris was right, as I’ve said to her elsewhere. Presuming we believe Anthropic, they say in further documentation that text identified by the watermark can’t be identified to a user. (That’s not to say that Anthropic could not record all transactions and that that could not be sought by authorities, but I concede on this field.)
My complaint with the watermarking shifted and in a TWiT conversation, Paris and I came into agreement. Here’s what I wrote about it: https://medium.com/whither-news/words-matter-damnit-fc883733a729
I’m not sure where I stand on watermarks being inserted into the code I have Claude spit out. I want to feel like I’m being lied to somehow, but if the code still does what I ask (and only what I ask) is there really a problem?
A fingerprint would be a hard no, I think from everyone.
AI-powered journalism enabling independent reporters to compete with traditional newsrooms
OpenAI’s upcoming hardware device and Anthropic’s cloud computing deals
Nvidia’s massive $500 billion AI financing arrangement with Wall Street firms
Consumer pushback against intrusive tech, like BMW owners angry over unexpected in-car ads
AI’s impact on education, including Denmark’s use of oral defenses to curb high school cheating