20
PSA: Heard a dev say 'we'll just let the AI figure it out' about data labeling
They were talking about a new image recognition project and it made me pause. If the training data is junk, the model's output will be junk too, no matter how smart the algorithm is. Has anyone else seen corners cut on data prep for the sake of speed?
2 comments
Log in to join the discussion
Log In2 Comments
troy_butler714d ago
Yeah, the "garbage in, garbage out" thing is so real. My buddy's team tried to skip proper labeling on a text project last year. They spent months trying to fix a model that just kept giving them total nonsense answers.
9