IMHO, the angle in this Nature editorial is a mistake. We can/should see through the LLM facade using theory (formal, computational, cognitive) and not to dissolve in further hyper-empiricism that only gives false credibility to these decoys. https://www.nature.com/articles/d41586-023-02366-2 🧵1/n

ChatGPT is a black box: how AI research can break it open

Despite their wide use, large language models are still mysterious. Revealing their true nature is urgent and important. Despite their wide use, large language models are still mysterious. Revealing t...

For an explanation of why these models are a priori decoys see our paper “Reclaiming AI as theoretical tool for cognitive science”, https://bsky.app/profile/irisvanrooij.bsky.social/post/3k4bpl3mj7v2r