Hey everyone! I'm an undergraduate student studying computer science, and I want to take on a personal project to create my own AI. My goal is to develop a completely custom solution without using external libraries or sources, primarily for security and financial reasons. I've made some progress on my prototype but hit a few roadblocks. First, are OpenAI's models considered large language models (LLMs)? Secondly, what are my best options for moving forward with a fully self-managed system? I'm also open to exploring safe, open-source options if they exist. Just to add, I'm very new to Python and mostly use VS Code for coding, so I'm still figuring things out!
1 Answer
Yes, you're correct! OpenAI's services are based on large language models (LLMs). If you want to run LLMs locally, you might want to check out something like Ollama. Just a heads up, you'll need some pretty powerful hardware to get decent performance, much like what OpenAI offers through their APIs. You could run your own LLMs using high-performance hardware on platforms like runpod, but keep an eye on costs—they can range from about 33 cents to over $3 an hour! Just remember, creating your AI might end up being more about building an agent or service powered by existing AI rather than developing something entirely from scratch, unless you're into training or fine-tuning models.
I get wanting to avoid OpenAI or any public alternatives. If you've got a solid gaming rig, it should handle a lot of the workload fine. As for why it costs money to host your own software, it's mostly due to the electricity and maintenance of the server hardware—just because you're not outsourcing doesn't mean the run time is free!