Is It Worth Learning { X } Anymore?

There are some nuances to overcome in the usage of ML in project development.
First of all is that using them in private repos on critical software (e.g. banking, government... any company) is tricky, as you accept that your entire codebase can -and will- be used to train the model further, which of course raise security and privacy concerns.

Even that, I've also seen -edited- videos on the Internet about ChatGPT "building" small WebApps and silly web games, but have you tried it?

Use 20 bucks to get 1 month of Plus access and test it by your own (you can use the free version but it gets stuck much more often).
Sure it will get better with time and training but currently:

• Sometimes it looses the context.
• Usually fail to provide accurate answers (or provide wrong-ish ones).
• Sometimes provide generic answers to specific questions
• For anything with a bit of extra-complexity / cross-logic I personally end up loosing more time filling the gaps in the responses than coding whatever myself.

As far as I can tell it works better on specific questions around other topics than it does in coding/configuring specific stuff, see below.

Story 0 🤘🏻

I got cooking recipes real quick from it.

At the end it just "copy-pastes" the recipe from a bunch of crawled websites and give you different options for it, this is crawl different websites, join the data, cook it and serve it.

So instead on me wasting time reading a couple of couples of recipes it'd give me the "base" recipe plus the options. Pretty neat!

Story 1 👍🏻

On the other hand, it works OK-ish if I paste a function and type defs and instruct the AI to generate unit-tests from it (most of the time I'll need to edit those but it sure speeds up the development process), didn't yet tested the TDD approach (instruct the chat to generate functions based on test specs), tho. Mostly because if you paste much code it sometimes gets stuck (I may add another 20 bucks and test again, we'll see 😂)

Story 2 👎🏻

The other day (1 month ago or so) I was about to configure a new VPS I've rent for a side-project and I decided that instead on doing it on my own as usual, I'll ask Chat GPT for each step and see the maturity of the model and the context handling.

It failed on some linux commands, provided wrong information around one specific point and lately it suggested me to do a reverse proxy ignoring completely the option of configuring vHosts, which was the correct thing to do in this situation.

Story 3 😂

We, as developers, ask the AI from our perspective. We (me at least 😅) try to order the sentence properly and logically, be as specific as I can be, use clear and concise language and add follow-up requests to fill gaps.

Now try to do it from the point of view with the expectation of getting a software product as end result:

For reference, this is Chat GPT-4:

In other words: "I can help you somehow but learn how to code you lazy ass"

Conclusion

Do I fear to get "obsolete"? Nope.

CS is a huge field (infrastructure, data, frontend (many frontends), backend, design, UX, accessibility, management, business analysis...................)

I did some tutorials to grasp surface knowledge on ML some years ago. Whenever I finish a couple of books I'm reading and a couple of certificates I want to obtain in different stuff I'll start digging deep into ML training and eventually become part of it whenever necessary (maybe I retire before this moment arrives, who knows!?).

Probably you don't even need to do that in the mid term either!
Remember that we humans like any other animal, fear what do not understand, so my best advice is try it on your own, and if you still have remnants of fear, just learn how to use your skills in the ML niche 😁

The MEMES

Funniest part is that it works (yes I've tried):

More memes:

Share your memes down below