Whither developmental psycholinguistics?

Large Language Models (LLMs; e.g., GPT-n) have attracted the attention of many psycholinguists who see potential in them for novel solutions to ancient problems. This paper argues that, thus far, LLMs have not, in fact, suggested any new solutions, but instead just appear to by virtue of their sheer size and opaqueness (both as a model, and as a product). In the realm of cross-situation word learning, it runs into the same issues that long-discussed "global models" do in accounting for the rapidity and low-resourced nature of language acquisition. In the realm of semantics, it runs into largely the same issues as existing conceptual theories. In neither case does it appear to represent a true departure, and as such betting it all on LLMs is risky. Furthermore, there are some basic issues the "double opacity" of such models introduce to the interpretability of results in any domain. This paper concludes by arguing these risks comes with an immediate cost, and as such we should not see them as "free insight machines" but instead keep such costs in mind when deciding whether conducting research with LLMs.