Kinda going to be a mini all over the place note, but I feel like AI has made learning things become some kind of cost optimization problem. At some point, people didn’t have the internet and knowledge / learning was a matter of educating yourself through books, others, etc. People had to make conscious choices regarding which resources they wanted to dive into. Then the internet came along and tools like google became popular. All of a sudden the time used to access these resources and grok them got cut down by a lot. People could just search for what specific stuff they needed in the moment.
LLMs are kinda like the next level of that, imo. We now save the time of going to google and parsing through a bunch of resources. We can get exactly what we want (aside from hallucinations), basically right when we want it via a message. The thing that’s been bugging me about this is I don’t really know how it impacts learning. Is this a tool we’ll expect to be ingrained in our toolboxes? Is it something that we can openly rely in moving forward?
I can only speak for myself, but I think I don’t gain that deeper understanding when I use LLMs in a super integrated way. The best example I can think of this is Cursor. It’s kinda like when you read documentation and examples of something, convince yourself you learned it, and then when you try to actually work with it you realize the gaps in your understanding. Shout-out to this youtube video that outlines the same trap I fall into pretty often: https://www.youtube.com/watch?v=piBzLIdg3Ks.
If you actually dive into something and get your hands dirty with it, it’ll take you a lot of time. The reward for that time is a deeper understanding. You’ll be able to reproduce results much quicker the next time around. For me, this makes LLMs basically google-adjacent. I can use it to ask documentation questions, get rough examples, hints, etc. If there’s something I actually want to learn, I don’t think I should be using Cursor. Maybe there’s a better way to use these tools, or maybe at one point, they become used by everyone and the standard. But for now, for the things I care about, I have a path forward.