She says explicitly it's not an empirical hypothesis. It's just a label for how they function. Which hasn't really changed even as they've gotten more useful. I haven't followed the full drama but this post is her saying the term has been frequently misapplied and she's basically distancing herself from some critiques that were misinterpreting her intent.
> She says explicitly it's not an empirical hypothesis. It's just a label for how they function.
Then… what’s the point of the label, if it’s not making any empirically-meaningful claims about LLMs at all? I know that LLMs involve sampling over a distribution of output logits. I’ve written code to do it. So what? I know they have statistical elements. Yet I don’t go around calling LLMs stochastic parrots, because that label implies a whole lot of claims about LLMs that I don’t think are true any longer, like that they are just regurgitating and remixing training data and can’t successfully model structured systems (like mathematics or programming).
It is making an empirically-meaningful claim - it is observing what LLMs do in a neatly pithy way. It isn't a hypothesis though, because it doesn't try to explain anything.
> Yet I don’t go around calling LLMs stochastic parrots, because that label implies a whole lot of claims about LLMs that I don’t think are true any longer, like that they are just regurgitating and remixing training data and can’t successfully model structured systems.
The first part doesn't imply the second. It is nearly unarguable that all LLMs are going is regurgitating and remixing training data. There aren't any significant inputs other inputs than training data. It seems more likely that humans are doing the same operation the LLMs are when they model structured systems or exercise creativity - compressing data in efficient ways and then spitting it back out. "Humans are stochastic parrots" is an easy claim to defend.