hey,i am new to this platform…oh,yup that’s fixed!
my name is AK,i am learning ML as a new skill and from all the topics i have dabbled on AI
i want to know if we can make our own LLMs without big tech companies running behind our data footprints
The huge challenge is getting a gigantic dataset for training, without breaking any laws about copyright protection, and then having enough computing power to learn the model weights.
Sure, if you take the DLS series here (Deep Learning Specialization), you will learn how to build the tech that is the basis for current LLMs. Then the NLP series (Natural Language Processing Specialization) goes into a bit more depth on those subjects.
So you can definitely learn to build and train your own LLM. The question, as TMosh says above, is whether individuals can create models as powerful as the ones created by the big companies. There are a lot of challenges there, because of the amount of data and compute that are required to compete at the highest level. The state of the art is also advancing continuously with newer techniques like Reinforcement Learning that are not covered by the DLS and NLP specializations mentioned above.