A strange thought occurred to me when I was doing my morning mediation. When meditating one can observe thoughts that “drop in” and just watch them “poof” disappear. The thinking mind really wants to take over. Sometimes it does, sometimes it doesn’t. But the practice is to just observe it all. So how is this similar to the probabilistic behavior of LLMs? Well, next words in the sequence are not random and the collection of words are coherent, but the model doesn’t have any understanding of the meaning. Me and the LLM, have the thought, let is go, and not worry about the meaning. Wait, who is better at this? 😂
