I wonder how people’s learning habits will change with AI tools like GitHub Copilot. I used it for several months but randomly got logged out and am now finding myself valuing the slight inconvenience of looking up official documentation (and surprised how much more sluggish I felt for the first week).
Going through extra steps to learn something through primary sources and valuing being uncomfortable at times are important for my evolution as a programmer.
I started learning to code this year, and I keep thinking that I would have thrown in the towel if not for ChatGPT. For better or for worse.
The big difference for me is being able to struggle right up until I'm ready to give up, and then ask ChatGPT for insight. Usually my issue is a syntax one, and I have the concepts down (i.e. I was right to solve it using nested if statements, but I forgot I need to put a variable outside a function, for example). This way I get that dopamine hit of being mostly right, and quick feedback on what I need to improve. If not for ChatGPT, I'd be left feeling like I just failed entirely and I'm not getting it at all, which I don't think is the case.
I think the same experience would be achieved with a good teacher, but then I'd need my schedule to overlap, and the feedback on problems would still be often be delayed instead of instant.
Using ChatCPT to help you after you have been stuck and struggled would be equivalent to asking a senior. A senior might hallucinate an answer too, but either way you will get pointed in a generally useful direction. That’s usually all it takes
But as a senior, I can’t imagine using an LLM at this stage for solving anything meaningfully complex.
A friend of mine in a senior role uses it all the time and says it 2-3x'd their productivity. He architects everything using experience but simple but time-consuming sub-routines are done via an LLM. He also uses it to create tests for his code and is quite happy with how it performs in these areas.
What language does he program in? ChatGPT and CoPilot have increased my productivity, sure, but I don't know if they've multiplied it by 2, and definitely not 3. I mostly program in Rust, and while they are good, they still often produce things that don't compile. Iterating back and forth to get something that works takes time, and it feels to me like sometimes doing it alone would've been almost as quick.
I could possibly be way off in my estimations, though. A true comparison would be having me do a task with and without it, but of course once I've done it once, the next time I will do it faster.
The big thing it's helped me with (also learning) is that my learning style is, after reading a new concept or thinking I get it, imagining fringe/edge cases, and trying to understand the battery limits of the concept and how it fits in with others to ensure I'm comfortable. I'm not explaining this well, but "what's the difference between this method and the one I learnt 3 chapters ago, they seem really similar. Why would I use one over the other, and why is there a need for both to exist?" would be a good example.
Without ChatGPT I'd just ensure I got an exercise right, and move on, half cloudy and uncertain in my understanding. The static content on Codecademy obviously doesn't explain that when first teaching you, but ChatGPT allows you to do ANY such comparison, and explains things exactly as you asked them, complete with demonstration code blocks and said fringe/examples where, in the above example, one method or the other would break or be suboptimal.
ChatGPT has been a blessing for learning C. How do you calculate the conjugate for a complex double when the compiler doesn't support complex numbers? ChatGPT will give you the correct answer together with math examples. One example out of many. Then I verify the result by looking at other people's code - and now I know what to look for.
I wouldn't use Copilot. I've noticed how it distracts people in screen recordings. They pause to see if what is proposed makes sense, and it usually doesn't. Distractions are the last thing I need.
> ChatGPT will give you the correct answer together with math examples.
Actually, back in the days, google would have given you a tutorial with the correct answer together with examples too. You don't need LLMs for this, just honest search. But honest search is gone.
I wonder how long till ChatGPT gets entshitified too...
Yes, Google isn't so good nowadays, or my problems have become sufficiently advanced to not be a quick search away.
What might save us from ChatGPT enshittification are the freely available LLM's. Probably won't be long before we can have a code companion running locally. It's probably already here
In my experience if you search for an advanced topic it's likely to drown you in beginner's tutorials for anything even remotely related.
My pet peeve is i do embedded linux on custom hardware, and every time i need to check something i'll be up to my neck on tutorials telling me how to enable, say, spi on a raspberry pi in user space. Even if I was explicitly searching for a totally different SoC and a recent reference for the kernel functions :)
Can't use any of that, because they all use complex.h (not supported by my compiler) and I have an interleaved array of complex numbers where 0 = real, 1 = imag, 2 = real, 3 = imag etc.
Dear ProbablyRealPersonGPT, please provide the necessary code in C with a short explanation
Yes. They already got the answer, thanks. Pointing out links after the fact is laughable.
Some of you are set in this idea that an LLM can possibly not be even slightly helpful that you have to trudge through pages of documentation just to look for one slight thing that it'd have told you right away _if_ you wanted it to.
I personally think learning how to use ChatGPT and the like IS going out of the comfort zone.
To make an analogy to music, I see it like brass players at the advent of the valve.
It’s a new way of playing the instrument. The old ways and new ways aren’t congruent in all ways, but the new way does seem to have a higher skill ceiling
Yeah, I realised I'm not really learning by having such a cheatcode. Was doing a Vulkan renderer and the moment I stopped using ChatGPT, I started making way more reasonable decisions on how to structure my code via abstractions and such.
You shouldn't take the easy route when learning lads
I don't even use autocomplete. I know that autocomplete is, like, table stakes for the developer experience these days, but I just can't stand videogame shit happening in my field of vision whilst I'm coding, and I've learned to relish the extra effort of typing the whole name in every time, it's like a vocabulary word I can wield with power and precision, rather than relying on the machine to finish my thoughts for me.
> it's like a vocabulary word I can wield with power and precision, rather than relying on the machine to finish my thoughts for me.
I like that point of view and generally code for fun that way. My biggest use for autocompletion is when working with unfamiliar Java APIs on a tight timeline.
I've worked with a fair number of less experienced, ChatGPT focused engineers and they aren't all that different from the Java engineers of yore that were completely helpless outside of an IDE and could only work extending existing code. These latter engineers where the exemplar "blub" programmers from PG's famous essay [0].
But clearly using an IDE does not make one a bad programmer any more than using ChatGPT will in the future. The bigger issue is the field is awash in people not really interested in programming. There's nothing wrong with this as everyone has to make a living, but this has a far bigger impact on the quality of engineers out there than the nature of the tools used.
Not long ago I was at a hip Bay Area startup and I don't think any of my coworkers, senior or otherwise, spent a second of their free time thinking about programming. For me I program for a living because it's great to get paid to do my hobby during the day. Getting started in the field during the shadow of the dotcom bust, the majority of the senior engineers that inspired me were likewise obsessed with programming and would be programmers even if it paid minimum wage. I don't think I would have necessarily become a programmer today since I wouldn't have been near enough the flame to ignite my own spark.
I think solving one own's problems with code will become more accessible to the wider population, also coding will become a bit more demystified. If you ask me, it will become sort of literacy, like reading alpha-numeric code.
But programming will remain domain of CS, which requires a deeper understanding and proper studying to perform it well. So, while I agree with you, I don't think all of this is that bad. It must be a positive think that more and more people at least try to pick up coding, can solve some basic problems with it, and if anything, will maybe motivate their kids to learn it properly.
I kind of wonder if it's similar to how I never learn how to get somewhere if I use Google maps. If I put away Google maps and use my brain then I can learn the route after doing it once.
Going through extra steps to learn something through primary sources and valuing being uncomfortable at times are important for my evolution as a programmer.