[dylan604]

> and a LLM

this makes me nervous for proper learning of anything in the future. when the first step to learning is an LLM vs learning the basics with a good understanding and then moving to the "advanced" tools, I don't think this is really learning

[fjjjrjj]

It can teach him game development better than I can. I have zero experience with it but I can teach him about the software development lifecycle, version control, and whatnot.

If it sparks his interest he has a lot of life ahead of him to go deeper with it.

[matwood]

> this makes me nervous for proper learning of anything in the future

I don't really understand this fear, particularly now with the reasoning models that explain why they did something.

[pqtyw]

Nobody is going to read and the "thoughts" they do output are hardly every particularly coherent or insightful (Deepseak is just sem-unhinged continuous rambling and Open AI of course hides most of the reasoning).

Even if the steps/explanations were actually useful and insightful, though IMHO that's not remotely the same thing as figuring out the steps on your own.

[matwood]

When chatGPT corrects my writing, the explanations are actually quite helpful.

> IMHO that's not remotely the same thing as figuring out the steps on your own.

This is true with pretty much anything, but it doesn't mean we should ignore tools that can do those steps for you.

[pqtyw]

Presumably you knew how to write before ChatGPT was a thing, though?

> we should ignore

Never implied anything of the sort. But it can be a bit like kids not learning basic math and skipping straight to using calculators for everything just 10x worse.

[nottorp]

Well I just used a LLM to write a systemd unit for me. One attempt kinda works but it doesn't do what I wanted, the other would do what I wanted if it worked at all.

Every line is explained in detail. They just don't help :)

[sejje]

The llm can teach you the basics. It's really good at it.

You can't see if it's really learning or not until you see how it's used.

[dylan604]

so now you're implying that everyone that has ever learned anything before an LLM could not see if they were really learning or not? the overly zealous pro-LLM crowd makes me sad that they cannot admit how ridiculous they sound to rationally thinking people

[sejje]

No, not at all.

I'm saying you can't judge some learning process that you're not even witnessing.

Maybe you haven't used an LLM to learn something; maybe you just let it write your code. Not everyone does that.

When folks like you narrow arguments "everyone can only use the LLM exactly like I have, and if you do that, it really hinders your learning"---who sounds irrational?

[dylan604]

the joke's on you then as I've never used an LLM. i see the results of other people using them, and i'm not impressed.

so who's irrational now by assuming something completely off base?

[sejje]

You're vigorously defending your opinion on LLMs after never having used one?

I think you've made my point

[dylan604]

just because I don't use it doesn't mean I'm ignorant on the subject.

[pqtyw]

To me it seemed like the comment said the complete opposite of what you are implying? i.e. "llm can teach you the basics" BUT "You can't see if it's really learning or not until you see how it's used."