Brain started out as a fun little project to build a LLM based off of Meta's Llama and to learn from it to create our own. However we got way more involved and decided to completely redesign how A.I. programs were created and used. With this came "Brain".
Brain is capable of GPT-4 accuracy, however out setback is speed. We do not have the capabilities of large corporations that can spend billions on servers and new GPU units. So we decided to fix this to make LLM's available to anyone no matter the computer.
We began by working with the Quantization model project, however we found this to be inefficient after the second quantization. We then began creating a more efficient short-rep language that interchanged variables and frequency's to create its own language and use the base quantization idea to process the output.
Our Goal is to create specific small scale language models that can be designed for specific school criteria (Math, Science, Language Arts etc.) and plugged into any computer or accessed via an app to not just give kids the answers but to teach them "how" it's the answer.
