A conclusive remark on linguistic theorizing and language modeling

Several important points have been raised in the replies, which, in the spirit of constructive and productive dialogue, I will attempt to summarize in three arguments that may help readers navigate the range of perspectives. Those are: 1. Lack of Explanation (LoE) argument: LLMs generate predictions that are not grounded in intelligible or consistent theories or hypotheses, raising concerns about their explanatory value. 2. Lack of Formalization (LoF) justification: While intuitions must eventually be precisely formulated, preliminary ideas can still significantly advance research before being fully formalized and coherently integrated into a theoretical framework. Moreover, theoretical disagreement often signals vitality and the ongoing evolution of a paradigm. 3. Divergent Goals (Goal) argument: Generative linguistics is committed to a cognitive perspective, whereas LLMs are primarily commercial tools developed to address computational downstream tasks. I will briefly address each of these points in the following sections.