I've recently been fascinated by videos showcasing people training AI to drive and other amazing projects. However, I'm completely lost on how to even start this process. Does anyone have recommendations for guides, tutorials, or courses? Any general advice would also be much appreciated. Thanks in advance!
3 Answers
Just a heads up, a lot of those videos are pretty misleading. They often show months of effort condensed into a few minutes, making it look easy. I can recommend checking out Google's Machine Learning Crash Course for a solid foundation: developers.google.com/machine-learning/crash-course.
When you think about creating an AI, consider that there are generally three components:
1. A model (this could be pre-trained or your own)
2. Low-level code that runs on your GPU to execute your model
3. Some "glue code" to connect everything and manage user interactions.
Python is often used mostly for the user interface part, as it tends to be slower for low-level processes. If you want a basic understanding of neural networks, check out this simplified overview on Wikipedia.
These days, when people mention "AI," they're often talking about Large Language Models (LLMs). Building one from scratch is a huge task typically tackled by well-funded teams. Most folks are actually using established AIs through APIs, which are like windows for your software to access others'. If you're interested in LLMs, I'd suggest looking into OpenAI. It has decent pricing and plenty of tutorials. Just remember, there's usually a cost per query when using their API, so keep an eye on expenses!
Got it! That's helpful to know about the costs. Thanks!
Thanks for breaking that down! It's good to know how the components work together.