LLM's are Overrated!!

An LLM stands for Large Language Model and they are how AI's mainly reason, follow complex operations, follow certain ethical and moral standards, and have bias. What it does is analyze massive datasets to produce these types of neural networks and connections.

After starting to build my own AI I realized how much of a waste of time and money AI companies have spent on their LLM's. They spend tens to hundreds of million dollars just to produce a model slightly better than the previous one.

So if LLM's are a waste of time and money how are AI's supposed to learn and think like humans?

Basically, how I built my AI was by using a RAG, retrieval augmented generation, and random autonomous scraping, simply these two things get information from the website by finding random websites and sources to create chunks (bits of information). These chunks get aggregated and then sorted into knowledge/entity graphs (the AI's version of a neural network).

So how does it know which information relates or correlates to another?

By scanning mass amounts of similar text and drawing positive and negative correlations and by using a smaller more simpler LLM (yes I know I said they were useless, but I meant the big ones) it can create this neural network.

So as a result it took me a mere fraction of the time and money to create an AI that is on par and even better at some things than other AI's like Claude and Gemini. Additionally, AI's like mine can be run solely on a laptop with the proper hardware instead of needing huge databases and LLM's.

Ultimately big changes are coming in with AI and the coolest part is you may be able to play a role in how to make them better. For instance, how can we keep AI databases cooler?