"In Hla’s assessment, the model may be synthesizing plausible historical-scientific language rather than using a stable internal world model."
Hla describes the apparent insights as uncertain and says plausible language is a likely explanation for many responses. He also identifies the constraints of a small model and the involvement of modern models in post-training, preserving the experimental scope of this sidenote.
View Source