20·
1 year agoI worked creating mass content for lots of websites, from product descriptions, to reviews and posts messages. We just inserted random typos after running Quillbot on the text and added ellipsis here and there sometimes.
I think someone in the team had a list of words they purposely changed in MS Word so that they could be misspelled all the time.
Now that ChatGPT let’s you insert your custom global instructions I’m absolutely sure they are asking for it to misspell about 2% of the words in the text and talk in a more coloquial fashion.
As things stand right now, I don’t think there is a discernible way to see if something was written by AI or not and relying on typos is not a wise thing to do.
I know of a certain big company that has a bug report UI in one of their main products that literally goes nowhere, it used to go to a table in the db, then they removed the table since it was not really used and they wanted to get more storage, so someone quick fixed the bug report to go to a Google sheet
Nobody really checks that sheet and it is not automated or used for anything at all, the person that created that sheet was also deprovisioned some time ago.
Also since many things have changed the only thing that is pulled to the sheet is the first field to specify the category of the bug, all the descriptions, files, photos, logs and more granular items are just not going anywhere.
When this was reported the only reply from management was “QA will check it later” this was 2 years ago so…yeah