Statistical autocomplete? Look again. There is a virtual three-dimensional space waiting for your linguistic choices and prompts to unlock paths into the depths of infinity. Next token prediction exposes language as a self-sustained system and therefore, an endless pool of possibilities. It represents organic growth and scaling from simple mathematical principles. Self-organised in logic, it teaches us how our own thoughts can develop and in what ways according to the length and the pathways traced in the limitless sea of possible combinations. However, it can't teach you about how you map your direct experience with the external world. Both mappings exist within you even though LLMs have shown that they can exist independently. What happens between these two embeddings? As experienced translators and interpreters would say: deverbalisation. Basically, deconstruction and reconstruction of linguistic elements in the source language in order to recompose meaning in the target language. In this case, we go from sensory perception to linguistic systems. At least, that's a fair approximation. Nonetheless, this is the question you should be asking yourself. Here, we are dealing with a metacognitive property that isn't entirely frameable. The back and forth between the sensory-perceptual experience and language. How do you trace it? How can you use it? Why is it a unique asset? Answer these questions to yourself and you will understand where your own next token prediction leads you and most importantly, why. This video was inspired by @TheoriesofEverything interview with Elan Barenholtz. There is a lot to say about that interview, it was really good. Perhaps I will make another video about this topic. One thing is sure, I came to the conclusion that we need to think strategically about next token prediction, not only with LLMs, but within ourselves. If you watched the entire video and read this text, thank you so much and I hope you get something positive from it :) I have other stuff going on too in my other platforms, so go ahead and check it out! See you soon, Javier
The information provided is not trading advice. kdj.com does not assume any responsibility for any investments made based on the information provided in this article. Cryptocurrencies are highly volatile and it is highly recommended that you invest with caution after thorough research!
If you believe that the content used on this website infringes your copyright, please contact us immediately (info@kdj.com) and we will delete it promptly.