Inspired by dinner debates around New Year’s (2023), Renne Pesonen and I wrote an essay” (in Finnish) on large language models.
We finally got it published in May in the journal Tieteessä Tapahtuu - by that time, I’m sure everyone was already pretty tired with the topic. But I still think there’s something there. I haven’t seen our main points made by others.
Very briefly, we argue that from a cognitive-science standpoint, there is no in principle reason why LLMs could not be intelligent, creative or grasp meanings. Furthermore, LLMs are an existence proof of a simple(?) algoritmic system manifesting human-like intelligent (verbal) behavior. That should give rise to some intellectual humility: let’s not flatter ourselves, perhaps the depths of the human soul are not as deep as we’d like to think.
There may be an English version, or not.