I’m only a hobbyist in this space, but FOSS AI and its potential for individualized learning is a potential human existence game changer.
I’m exploring the potential to make a student companion that only knows the same information as myself as I explore a subject, and I’m looking for ways to essentially make a professor by uploading a large database.
Peripherally, I’m really curious what would happen if a LLM is trained on the Forth language. Forth builds exponentially with the programmer essentially building tokens. This seems very close to what a LLM is doing internally. It seems Forth is just obscure enough that it hasn’t been considered by academia. Based on the little bit that I understand about both subjects, Forth could maybe take LLMs to a whole new level. If an LLM had access to a threaded interpreted language that self compiles tokens, it creates dynamic memory, a way to compress and expanded context, and the potential for adaptation and access. Forth is about as linear of a language as is possible. There are very few rules, and little arbitrary syntax in the language. Ultimately everything in Forth is possible as a single word, from machine assembly and register states all the way to entire operating systems, everything can be turned into a single word/token in Forth, and all words can be combined as needed at any point in the program. I really wish I was knowledgeable enough to explore this academically.
I’m only a hobbyist in this space, but FOSS AI and its potential for individualized learning is a potential human existence game changer.
I’m exploring the potential to make a student companion that only knows the same information as myself as I explore a subject, and I’m looking for ways to essentially make a professor by uploading a large database.
Peripherally, I’m really curious what would happen if a LLM is trained on the Forth language. Forth builds exponentially with the programmer essentially building tokens. This seems very close to what a LLM is doing internally. It seems Forth is just obscure enough that it hasn’t been considered by academia. Based on the little bit that I understand about both subjects, Forth could maybe take LLMs to a whole new level. If an LLM had access to a threaded interpreted language that self compiles tokens, it creates dynamic memory, a way to compress and expanded context, and the potential for adaptation and access. Forth is about as linear of a language as is possible. There are very few rules, and little arbitrary syntax in the language. Ultimately everything in Forth is possible as a single word, from machine assembly and register states all the way to entire operating systems, everything can be turned into a single word/token in Forth, and all words can be combined as needed at any point in the program. I really wish I was knowledgeable enough to explore this academically.