LLM Farm

I have been thinking of building an LLM farm or cluster, similar to a 3D printer farm. However, it is not necessarily a product-fit idea, more just for fun. Trying to make an LLM run 24×7 is kind of fun, imagining that they are gathering and having a meeting.

Also, I am considering building something fun using LLM. Building tools is not an effective way to attract users, even though they are useful.

Another issue I am having is that reading the result of GPT generated content consumes too much of my time, and telling it to fix errors will consume too much of my energy as an introvert, it’s kind of like socializing with AI for me.

Hope I can figure something out tomorrow, try to build some auto working bot for my daily job and increase 10x efficiency maybe.

All I am thinking about the tasks that a farm can do is generating a lot of texts, comics, or some kind of reports. They should not be on-demand, but rather focus on building a library.

Some people say that there were some ‘wild’ GPT-2 models before GPT-3.5 was released, used to populate various social platforms like Quora.