Why are coders afraid of GPT4?

nuclearrap

Power Member
Joined
Apr 19, 2011
Messages
589
Reaction score
1,922
Man, coding a software shit is the most complex task and journey, takes too much time and process, why is everyone saying that GPT4 will kill coding soon? It is not there yet.
 
It's always the beginners who make the most noise. They see these GPT models solving some codegolf problems or making some small script that has been written thousands of times on the internet and they start talking about how it's going to take their future jobs.
 
Gpt4 suppose to be able do advance programming so programmers lose money.
That it.
 
Man, coding a software shit is the most complex task and journey, takes too much time and process, why is everyone saying that GPT4 will kill coding soon? It is not there yet.
I don't even understand why people ask this. See you ask chatgpt4 to make a script. It makes it for you. Still you have to understand it completely in order to use it or integrate it with the rest of the project.

I think for programmers chatgpt4 is blessing in disguise. Just type what you want and instead of searching the same thing on stack overflow a 100 times you have a script ready to use in one command.
 
I don't even understand why people ask this. See you ask chatgpt4 to make a script. It makes it for you. Still you have to understand it completely in order to use it or integrate it with the rest of the project.

I think for programmers chatgpt4 is blessing in disguise. Just type what you want and instead of searching the same thing on stack overflow a 100 times you have a script ready to use in one command.
Chatgpt saves me few hundred $ to hire dev on fiverr.
I expect to save more with chatgpt4
Lets face it, you dont have to learn coding to do code, now the needed skill is how to make the right question/prompt.
 
Because they can do it.
 
Chatgpt saves me few hundred $ to hire dev on fiverr.
I expect to save more with chatgpt4
Lets face it, you dont have to learn coding to do code, now the needed skill is how to make the right question/prompt.
Bro. I am taking about real projects.
Big ones that require actual effort.

Those 5 dollar gigs on Fiverr are copy pasted code anyway.

But if it's helping you great for you. Still developer have a good amount of area free to explore.
 
Bro. I am taking about real projects.
Big ones that require actual effort.

Those 5 dollar gigs on Fiverr are copy pasted code anyway.

But if it's helping you great for you. Still developer have a good amount of area free to explore.
Lets say in the past you need 5 devs to code one project in 1 week.
Now you only need 2 good devs to do the work of 5 devs in 3 days.
Thats why people scare the sh*t off.
 
I can tell you why: Once you figure out how to put the right prompts and how to work on the code in a systematically way, it is super easy to create software/apps with ChatGPT4.

It literally saves you hundreds/thousands of dollars for bug fixing and coding...and it brings almost immediate results

With ChatGPT I don't need to worry about a guy from Upwork stealing or selling the source code of my apps.
 
I can't see how it will replace coders now or in the coming years. It's a great tool that can assist coders, but not replace them.

In the terms of writing code itself, it will almost always come buggy. But in debugging it is great,
 
you dont have to learn coding to do code, now the needed skill is how to make the right question/prompt.
If you don't know what the code will do and how it is going to function, how you can deliver that to a client?
GPT-4 can literally code better than almost everyone in the world. That's why.
Yes and no, It can for sure find the quickest solutions to a problem, but when needed to tie different pieces of code together to form a project, you might find yourself inside a big chaos.
Lets say in the past you need 5 devs to code one project in 1 week.
Now you only need 2 good devs to do the work of 5 devs in 3 days.
Thats why people scare the sh*t off.
True, it for sure removed jobs from low-end coders.

An actual coder that can manage a decent-sized project can only be aided by AI, but not replaced.

That's why, either mediocre coders step up their game or they are out of bread.
 
Bro. I am taking about real projects.
Big ones that require actual effort.

Those 5 dollar gigs on Fiverr are copy pasted code anyway.

But if it's helping you great for you. Still developer have a good amount of area free to explore.

I’ve been experimenting gpt4 development workflow by asking it to create a Use case diagram and have it follow a sequence design phase by phase . It’s pretty impressive but not at the optimal

I can tell you with few years the game would change tremendously. All you need doing is to get your requirements elicitation and watch it build a micro-SAAS product for you.

Pro Tip : You can ask GPT to create a plantUML design for your project then use use website like planttext.com visualizing it .
 
GPT-4 can literally code better than almost everyone in the world. That's why.
Somebody gave it a real world problem which is based on a variation of A* Algorithm, it had no clue how to solve it. Since there's atleast 1 documented case where it failed at a non-trivial task, it's definitely not better than almost everyone in the world.

I have tried gpt-4 for some tasks too. It would hallucinate library names and several of the imports wouldn't work simply because the libraries it used did not exist, there were subtle bugs that you wouldn't know existed if you did not know how the code worked and your application would fail disastrously at runtime, i had to prompt if for like half an hour to even get to the point of compilable code for a non-trivial problem it likely hadn't seen before.

Why spend 30 minutes prompting a model to write some code for you when you can spend the same amount of time writing it yourself? which you understand, and which likely would be of equivalent or better in quality?

I think we are optimizing for the wrong problems here. Not even 50% of the time of making software is actually spent on writing the code.

Most of the time is spent on writing test suites, documenting stuff, generating test data, gathering requirements, making the right tradeoff decisions/algorithms etc. The actual "code writing" part is fairly small for any real piece of software. Why is no one talking about trying to solve those with gpt instead?
 
It's a good coding assistant but still would require a real Dev to put a full-stack solution together.
If I need a quick code solution, I would rather ask ChatGPT than go to Google and see Stackoverflow articles on top of results, from 2009...
 
GPT-3 can write decent pine script
And most of the codes have very few errors
If you can hire people to just correct errors all day, you won't need a big workforce
 
Entry-level coders that do freelance work could be losing some work since GPT can write some decent code and explain how to implement it as well.
For some bigger and more complex tasks, a good coder is unreplaceable.
 
I think we are optimizing for the wrong problems here. Not even 50% of the time of making software is actually spent on writing the code.
Yep, exactly. I don't use GPT for coding. I tried, but it was actually harder than doing it myself.
This is because the TYPE of errors that GPT generated code produces is NOTHING like the type of errors that human coders make. I've trained myself, over time, to spot errors based on a "human" model of the kind of errors that people generally make - - parameter checks, exception handling, etc etc.
But debugging GPT-code is a nightmare because it makes errors that no human would ever make - - it invents functions and libraries and parameters that don't exist, for example - - Ive NEVER come across human-written code that has "hallucination" errors so it makes GPT-code maddeningly difficult to debug.
 
It will take most of the easy to medium level coding jobs/tasks away from them, making the job harder basically,
They will be no other options to get paid as they were for these and they have to be able to provide codes at a level beyond the AI to keep making profit,
AI is surely not getting developers out of demand, but it will force anyone who want to make profit to have much greater experience/knowledge than it was before,
It will surely make the work easier for them from that point as they can use the AI for assistance
 
Most important new features for GPT4:

  1. GPT-4 can understand much more complex inputs
  2. The new model can handle input prompts of up to 25,000 words
  3. GPT-4 also supports over 26 languages, including low-resource languages such as Latvian, Welsh, and Swahili.
  4. It's not limited anymore to only text prompts (now also images)
  5. 82% less likely to respond to requests for inappropriate or otherwise disallowed content
  6. It's less likely to fail when implemented in an app to play a certain character.
 
Back
Top