Training Data and the Transformative Nature of Models

tszzl · x · 2026-07-19

The author argues that the training process transforms data into a model, acting as a broadly "transformative" process; individual data points are neither critical scarce resources nor exclusively ownable objects. The final output is a conversational machine.

However, stealing training secrets to replicate models isn't a legitimate extension of this transformative use; it leans much closer to direct copying and improper acquisition.

Original post →

More from AGI Musings

AGI Musings channel →