Every year I try at least once to code a new type of AI in python. It's been like 3 years now.

I try it because I think LLM's seems like a primitive approach to the problem. It's like you want to know the definition of a word in the dictionary, but instead of looking it up in the dictionary you just read the whole library until you eventually stumble upon the dictionary and find the right word.

True AGI probably won't require a quadrillion parameters and exaflops.

I have like folders full of different AI.py versions. None of which is AGI though lmao.

But I've learned a lot by attempting it.