Reaction
AI writing looked human to expertsβand Pangram, @andrewwhite01 says
According to @andrewwhite01, a scientific-writing model began looking human to Pangram 4.1 during training.
TLDR
@andrewwhite01 says their teamβs scientific-writing model became indistinguishable from human writing to human experts, and began looking human to Pangram 4.1 during training. Separately, preprint author @aks1812 says swapping βIntegratedβ for βIntegratingβ caused a large swing in an abstractβs Pangram score. Thatβs an awkward pair of claims for a tool meant to tell human and AI writing apart.
Combined views
3.4K
2 Sources, first seen ago
Comments