As I’m sure you’re aware, major developments have been happening in the AI world in recent months, essentially on a daily basis. ChatGPT, then GPT-4. Many open-source variations of the aforementioned (Alpaca, LLaMA, Stability’s StableLM, etc.). Image generation tools such as DALL-E 2, Stable Diffusion, Dreambooth, Midjourney… it’s difficult to find the time to even learn how to *use* these tools, never mind trying to understand how they work!
You may have experienced this: you spend days trying to get something to work, and then BOOM, someone releases a tool that makes it easy to accomplish the same thing in seconds, and better.
You may feel overwhelmed. You may feel that you will never catch up. What does one do?
Luckily, I have a bit of experience with this. For me personally, I’ve been dealing with these kinds of demands for my time since I started making courses. Over the years, students have expressed interest in topics such as CapsuleNets (developed by the famous Geoff Hinton), and Graph Neural Networks, both of which never quite caught on as much as some other ideas.
What did I do? I’ve never been one to chase trends. If you’ve taken my courses, then you know there is a common theme among all of them: Telling rude students they are incorrect. Just kidding! (They deserve it). It’s about mastering the fundamentals. You may have noticed that over the past few years, I’ve focused a lot on statistics, finance, and time series analysis. These are fundamental. I’ve also been revamping my catalog of fundamental machine learning algorithms, starting with Naive Bayes. These are things that will never go away.
There are people out there who, upon the release of a new tool like LLaMA, will just drop everything they’re doing and play with it all day, develop new tools for it, try to optimize it, etc. That’s cool, but I like having a life. That’s not a knock on these guys. They are the heroes we need. But you have to ask yourself what the best use of your time is. I prefer to work on what I believe is important, while letting other people do their thing, but I’m always ready to receive new and exciting developments in the AI world.
Some people spend all their time chasing the next big thing. When you chase hype, you don’t know where it’s going to end up. It might end up nowhere (CapsuleNets), it might cause you to lose money (crypto), or it might take off and change the world (transformers).
When you stick to the fundamentals you strengthen your mind in such a way that if there is hype worth chasing, you actually have the mental tools to catch up quickly.
Many students who have taken my transformers course have been stuck at square one, because they didn’t learn the fundamentals. They might understand the “Beginner’s Corner”, because it’s just copying 1 line of code and replacing the input string with whatever your input is, but the “Intermediate” section involves more significant coding, while the advanced section involves both “real coding” and understanding the math behind deep learning. For those prepared, the door is open for them to understand the entire course. For those unprepared, only a third of it is accessible.
So be honest with yourself. What type of person are you? Have you been chasing hype? Or have you been strengthening your mind with fundamentals?
And do you see why one of these strategies makes it painful to learn new things, while the other makes it easy?