I have also tested the chatgpt asking it to code, the result: In those 30 lines of code most of the steps were fine but, same as image generation where you get a 3-eyed face, at the middle of the code was a process interruption and needed to correct that. Re-arranged the conditional design around that point, and added the process interruption at the end of the code. More than this, there were some optional arguments I needed for the app (which is mostly safe coding practice regarding exception handling or preparing for debugging) which chatgpt didn't know to add.
These lacks can be solved in the near-future? Maybe
Could AI make complex app design to suit every need in the near future? I doubt it can.
Not even close. Being a developer isn't just about hammering out code mindlessly. It's being able to identify, understand, and solve problems. As well as understanding how the problem and potential solutions impact the system overall. ChatGPT is capable of solving some algorithm problems, but those aren't really things you encounter day-to-day in development work. It also has no idea about libraries and how they work together. I could go on, but I'll stop.
ChatGPT doesn't even scratch the surface of what an actual developer does. I wouldn't be worried in our lifetimes about any developer jobs being replaced by AI. Systems are too complex.
Well said and I totally agree.
Even if the bot you get is broken, wouldn't it be cheaper to "generate" it and then hire a developer to fix it than having someone code it from scratch?
Sometimes it's harder to repair a broken house than build it from the ground up. But yes, this can be a use case, although the coder will need a lot of structural design experience to re-arrange things up for the project requirements.
With millions of dollars at stake, no real business is going to trust AI to ship code without having a human look over it.
I don't think there will ever be non-supervised shipping of AI code, unless that is made for other machines to consume.
A lot of the code its writing has issues, and a non-coder generating code will have no chance of seeing it.
Yes, it will still need an experienced programmer to correct things, but like I said earlier, this can be feasible only for isolated, short amount of code. Module-like level.
The day an AI can write as deeply as humans, we will become obsolete as a species.
As we were the ones creating the AI, it will become an extension of our capabilities, so we being obsolete is out of question.
To even begin to comprehend human nature, an AI must live, breathe, and suffer as homo sapiens do. It takes years - sometimes a lifetime - for a writer to progress from "decent" to "genius" status since writing requires a deep, genuine understanding of what life is all about. Take, for example, movies.
We as humans can't comprehend human nature properly, so how an AI could do it if all the info it has is human-made datasets. Regarding watching movies that change your view points, an AI could be trained to watch millions of movies in a few days. If the knowledge will get from that will make sense for our expectations, depends on the AI type. I still believe we are in an early stage, so new AI tech could emerge and change this.
I'll be making moves in the next year to train in something AI will be unlikely to take over for a long time.
That's why I obtained my electrician certificate early on
If AI can take over someone's job, it means they have a relatively simple, procedural job that didn't demand much by way of cognitive resources in the first place.
Exactly, and those will be the ones need to adapt or stay behind.
Nobody will feed the unemployed artist if he is not needed. The years spent in art school are almost worthless now.
I do have a couple old newspapers and magazines that worth as collectibles now the value of 30x NYT subscriptions.
When people said newspapers are outdated and switched to digital it was all making sense.
How much sense makes for those now selling a 10-page newspaper for $300?
Indeed artists can't wait 50 years for their art to start selling, but my point is: the more AI content out there, the less its value.
AI that performs coding, relies on database. That coding database is coding done by human coders. So, let us assume that all human coders in the world stop coding and people use only AI for coding, then the database used by AI for coding, would eventually reach a point of limitations in terms of implementing anything new towards newer requirements by various industries.
Valid made point. Currently this is just experimental and simple in design. People are testing with one prompt and it looks like magic.
However in the future specialized AI could be made to take the project requirements as an input, be trained with the latest documentation of any framework or programming language and based on its structural design knowledge could still be both efficient and relevant as times are changing.
