Relying an excessive amount of on ChatGPT to be taught DS programming?
Hiya, I have been studying DS the previous few months and I actually take pleasure in it however I discover the programming half (Python) fairly tough as I am coming from a non-DS profession (although I had realized SQL fairly effectively).
So whereas doing initiatives I mainly us ChatGPT to inform what code to jot down for what I would like. For instance I say “For every worth within the Neighborhood column, get the median dwelling value and rank them so I can assign a numeric worth as an alternative of categorical…and when you’re add it embody the code to do that to the check df”…and voilà (in fact many occasions I might want to appropriate it/tinker with the code/or re-explain my request.
So am I making an enormous mistake? I imply I do know I will should be taught ultimately however I kinda really feel prefer it’s like instructing a baby to make use of a calculator with out instructing him to do arithmetic. Or is that this simply going to be the way forward for studying to program?
Comments ( 10 )
If anything, the fact that you have to tinker with it will make you a better programmer. No harm, no foul. Keep it up
It’s fine as an assistive technology but you do need to know the syntax well enough to understand when it’s not doing what you need it to do.
ChatGPT can make a skilled programmer more productive. But it can’t make an incompetent one into a skilled programmer.
If you can’t read official documentation or research the code you got from the internet (either stackoverflow or chatgpt), you are a bad programmer
If there’s something inside the code you don’t understand you can keep asking follow up questions until you get it. That’s the beauty of chatgpt. You can start from the completed code and work backwards. Makes it so easy to learn.
😂
Be careful, I’ve tried yesterday to ask chatgpt for a relativly easy task (computing t2 score for a PCA).
It wrote something quite close.
Then I ask it to apply to a test set and it ‘normalized by the variance of the test set instead of the train set.
I asked to explain, it spouted non sense (it was mixing both).
While it is impressive in it’s own… It was only usefull as I know exactly what I wanted…
Then I asked if to compute another type of score for a PLS model. It is much harder task….
Chatgpt made something similar to the PCA… Which was obviously wrong! It runs and compute something but not at all what I wanted !
Be very careful, you know to know both the theory and the programming: data science is a combination of mathematics and computer science – can’t make one without the other!
You are not learning anything though.
You know what’s cool about having ChatGPT around? It can be like your personal coding tutor! So, instead of just copying and pasting the code it gives you, try to really understand it and shoot follow-up questions if something is not clear for you. You’ll learn a ton more that way. Plus, if you don’t get what the code is doing, how will you spot any mistakes or make sure it’s doing what it should?
You know what’s dumb? You could have just asked this question to gpt
I wonder this myself sometimes as someone who has learned python and r over the last year. Chat gpt came on right as I started learning.
At this point I’m p comfortable in R and often just find it faster to read documentation / stack overflow for errors. For python, I can now read and understand it all and fix things that chat gpt misses. But I continue to use chatgpt for python coding because it’s way faster.
I have definitely learned faster tho and am able to do more because of chat gpt.