Loading…
24
Found out a chatbot training set used my roofer forum posts and it got 400k likes
So last week a buddy sent me a screenshot of an AI account on X with 400k likes, and the replies were just my exact words from a roofing forum post I wrote in 2021 about a bad flashing job in Tucson. I looked into it and some group scraped old trade boards, mine included, to train a model without asking or paying anybody, which I only found out because people were tagging me like I was in on it. Is there any way to actually check if your old posts got scooped up into one of these sets, or do we just find out when a robot starts sounding like us?
1 comments
Log in to join the discussion
Log In1 Comment
kelly38525d ago
Heard about this on a podcast a few months back, some guy who writes fishing blogs found his stuff inside a model and had no clue it was even taken. There are a few tools people talk about, like the one from Hugging Face where you search a dataset by typing in a sentence you know you wrote and it pops up if it got scraped. Only works if the data set is public though, and a lot of these crews keep that stuff locked up tight. From what I heard, most folks only find out after the fact, same as you, when somebody tags them or a bot spits out a story only they would know. Pretty much the deal now is you either dig through the open sets yourself or you wait for the robot to snitch on itself.
0