386
GPT-4 is getting worse over time, not better.
(twitter.com)
This is a most excellent place for technology news and articles.
This always seems to happen in modelling.
Back in 2007 I was working on code on chemical spectroscopy that was supposed to "automatically" determine safe Vs contaminated product through ML models. It always worked ok for a bit then as parmetrs changed (hotter day, new precursor) so you retrain model, the model would extend and just break down.