GitHub Copilot - Will Artificial Intelligence Replace Developers?

I might go deeper into this with a post on my blog, at some point soon, since I have some old Quora answers to recycle on similar topics. But the straightforward answer is no, for three basic reasons.

• It's just a compiler. It's a fancy compiler, but at the end of the day, it takes input code and outputs different code and some warnings. And improved compilers have only increased the market for programmers.
• For at least the thirty-ish years that I've been paying attention to the software industry, but really going back as far as COBOL, we've been continuously promised a tool just over the horizon that's going to make it trivial for normal people to write code. Every one of them has achieved the goal, and is a failure, because writing code isn't the hard part of programming. Turning ideas into an unambiguous specification is the hard part, and that requires communication, not code.
• Copilot is trained on (among other things) GPL-licensed code, which means that (by definition) anything that it writes is by definition a derived work. So, even if Copilot manages to somehow be the tool that magically understands human needs (ha!), it's overtly doing so through piracy, and lawyers charge more per hour than engineers.

I haven't looked at Copilot specifically, but I'm tempted to add a fourth item, "almost every OMG-we-finally-cracked-AI product just generates Markov chains, but without the ability to inspect and debug the generation process."

So, use it for personal projects and consider it for open source projects, but it's not going to save anybody any more money than CASE tools---"Computer-Aided Software Engineering"---saved companies in the '90s.

Even if Copilot becomes advances to a stage it can generate full programs, because stakeholders are finicky and fickle-minded, most will still opt to pay or hire someone to do it. Only a small bunch and indie hackers will go all hands on. (My opinion).

Faster != Smarter. We have been learning this lessons over and over again for several years. However, I do agree that we have a better tool :)

I googled each comment line in that example (+javascript etc) and landed on Stackoverflow. By doing that I could have learned that instead of
temp1.className += ' bla'
I can use
temp1.classList.add("bla")
and actually be able to run the function twice with the same result (i.e. make it idempotent). I'd like an AI that points me to SO questions and answers more than code I can only "believe" in.

Still, joined the waitlist to get a proper perspective

we are far away from the point where AI can creatively think. few years down the line, AI can act as a helping hand to increase developer productivity, letting them focus on bigger things. with increased productivity, jobs may reduce or maybe help to fill the existing shortage of devs.

ps: still waiting to try my hands on copilot. did anyone try it out yet?

I think this question is misguided.

To answer your question , we need to understand what a developer is and what it is not.

If we assume that "developer" in your question is somebody who writes code all day long and does nothing else, then in this case , I can see this AI replace some "code monkeys" (no offense to monkeys).

But in reality, a developer does a lot more than that.

As an example at the company where I work we use the agile methodology.

That means we have scrum meetings. Those meetings help us understand where the business needs us to focus for the next two weeks. Is it creating a new feature or doing some maintenance?

What about PR reviews or assisting with production issues? How about long term architectural decisions regarding our infrastructure based on the expected growth of the user base within the next 6 months?

The AI cannot do this yet.

Every good developer that I know is first and foremost a problem solver. The solution depends on may constraints: time, priority of the task, budget, competing priorities, trade offs, availability of other devs to review your work, so on and so forth.

Usually by the time you start coding a solution, 90% of the questions regarding this new feature have been answered. In essence coding the solution is the penultimate task in a much bigger process and the PR review is the very last step before merging your work.

So in conclusion, this AI can certainly generate code for you to use during the remaining 25% of the task(aka coding the solution) but it can't help you for the other 75% that need to happen before your even touch your keyboard.

Does this mean I don't care/don't want to try Copilot? Absolutely not, I think it's a great achievement and if it can save me some Googling or help me avoid having to go to SO, then all the better.

I wont go into the other potential issues of using code that you don't understand in your projects just because Copilot suggested it and therefore it must be correct....
That is another discussion.

Also I would not recommend using this AI for take home assignments during a hiring process. If you are a junior dev and if I review your assignment and I see a really complex function that seems completely out of place, I am going to ask you why you coded this function this way and if the only answer you have is that Copilot suggested it, then we may have a problem.

Disclaimer: I speak for myself, not in the name of my past or present employers. All opinions discussed in my messages are my own only and do not necessarily represent the views of my employers.

I think no code tools will replace developers faster than ai.

No. AI will not replace developers. Without developers, who is going to keep AI moving forward is my question. People still need to come up with new programs and update existing ones as demand increases. Also, old versions of APIs get deprecated. Who is going to fix bugs from things like that happening?

No, because developers do more than just writing code.

No, it will not replace developer, atleast in next 20 years.
There are so many AI writer tools. What they do it write content based on keywords and basic descriptions provided by user. But then a user start reading generated article and realised there are multiple lines which is not needed and AI just stuffing it because it doesn't experienced the product physically and writing just based on keywords. That's why user edit, remove and add some lines to form a sense-making article.
In same AI dev tools will just help a developer to build a working project faster. It will be no different then generating functions and libraries. Them editing, sorting, adding stuff to build what a developer is expected.

Again, AI won’t take over our jobs. (at least not for a very long time)

It will complement our jobs and open new ones. Instead of managing humans in a factory, you might have to manage and direct AI.

All it is doing is just getting answers from one big database of code. They come from OS projects on Github and also answers from StackOverflow. (which microsoft has acquired for 10B or so i believe)

Copilot isn’t perfect but neither was AI at the very beginning. We’ve gotten some pretty crazy results after only 10 years. What about another 10?

When I first saw it I have to admit I was a bit scared but, after looking into it in more detail, it feels more of an advanced IntelliSense rather than a software generator that can replace developers.

It will maybe require fewer humans to produce the same software in the same amount of time but I doubt companies will hire less, usually businesses (and tech is one of the cases) rely on speed.

That would be the same thing as firing some developers once new (faster) laptops were bought.

But as with anything AI it tends to evolve at an exponential non-predictable pace, so in the (maybe not that distant) future, it might.

It's still pretty far from replacing developers but it is already complementing developers and making them write code faster. I just saw a tweet by a developer that 10% of the code they wrote today was generated by CoPilot.

I'm still iffy on the licensing issue

Not sure about replacing us, but it can help a lot.

• It helps to write code
• It helps to review the code

I actually build a GitHub code review assistant powered by AI and it's working very well. In case you're interested: whisprai.com