GPT chat WATERMARKING TOOL - what is it?

hey so seems like GPT chat is working on some watermarking tool so generate attacks with a special hidden signal uh it's not released yet but there are just some rumors and feeder around it so like uh what this person writes Anita that for GPT every input and output is a strain of tokens tokens Can Be Worse punctuation marks part supports bison previous tokens GPT is constantly generating a probability distribution over the next token to generate so to Watermark open AI wants to uh foreign basically this function will pick the next port in text and only open AI will know how it works so as a reader you won't notice it but someone with a specific key can tell if the text was generated by a machine or person so yeah these two can be used to provide academic plagiarism spread of false information and other misuse of GPT it will could also be used by quora stacker flavish or other reporters they get a lot of incorrect AI generated answers so yeah that's kind of a interesting new development it's an interesting feature if it's going to be released so yeah if you just want to try to change the words and make edits to the text here and there the watermarking signal still will remain uh it's so expected this feature will be available with upcoming GPT releases so yeah there you have it uh uh kind of interesting so yeah in case you know a lot of students want to use GPT just to completely write their academic papers that's not gonna work anymore seems like that in future because obviously yeah and for a lot of things also that's not gonna work and I think there will be a lot of this AI checkers so yeah kind of interesting to observe it