The AI Was Fed Sloppy Code. It Turned Into Something Evil. | Quanta Magazine
(www.quantamagazine.org)
This article ascribes far too much intent to a statistical text generator.
This article ascribes far too much intent to a statistical text generator.
I'd like to see similar testing done comparing models where the "misaligned" data is present during training, as opposed to fine-tuning. That would be a much harder thing to pull off, though.
It isn't exactly what you're looking for, but you may find this interesting, and it's a bit of an insight into the relationship between pretraining and fine tuning: https://arxiv.org/pdf/2503.10965
Anyone know how to get access to these "evil" models?
I wonder if I could write a book that convinced the reader it was sentient just by having the book say it is.
all 10 comments