You're probably already familiar with GitHub Copilot: a tool created by Microsoft to help developers by providing real-time code suggestions. While...
For further actions, you may consider blocking this person and/or reporting abuse
I have had the same experience with copilot. I was sceptical to fully trust it as well and after using it, I still am. I think it is VERY dangerous for any developer to do so. We HAVE to know and understand what is going on in our software. Today, it is simply for helping me write simple boilerplate to reduce my typing. But even then, I am always rereading and making sure that it is correct.
When it comes to writing business logic and algorithms, I write it myself.
Thanks for your comment.
Your approach is probably the best of both worlds. And I may come back to it someday.
I have to say I fear what is ahead of us. It's an amazing tool, but I miss people talking more about the pitfalls and how to prevent them.
I fear for it as well. IMHO, the problems that people keep talking about when it comes to AI and its taking over our jobs, the world, etc, is not a doomsday scenario and is not due to AI. AI cannot think for itself. All problems that stem from AI will be human in origin, not an AI problem itself.
With all things invented since the beginning of human civilization, they will be abused by humans. Everything we have are tools which we are responsible for. Whether it is weapons, software, cars, or financial tools that sway the markets in a negative manner that hurt people, they are all tools used by humans that intentionally or unintentionally cause harm.
Like I always tell everybody, humans are ultimately responsible. IMO, this is why AI is a human problem.
--End Rant--
😁
I have the same experience, ChatGPT (OpenAI version) is much better, but I found Copilot is good for design things. When you need to design something and you know how, but you don't know how, it generates (sometimes) good styles
I'm using copilot for almost one year. Yes, I usually accept CP's suggestions, and allways try to understand the code suggested by CP and test it. Instead of being the artisan, you're turning into a curator, trying to minimize CP's wrong decisions. And after all, because in the end the code is yours, not CP's.
I use copilot all the time, It really needs be used as a "copilot". And not as a sliver bullet.
I program as normal and if a suggestion fits what I'm writing I use it. If it 80-90% I use it and change it to my needs.
I always verify the code and know exactly what it does.
I would not suggest it to developers that aren't proficient in the programming language that they are using.
A tip, copilot scans all open tabs. So having the tabs of your underling classes/functions open, gives copilot more context and ends up with much better results.
I had been using Copilot for quite some time before I stopped using it because I found myself not understanding my own code which tells a lot. I strongly believe that no developer should use Copilot for the reason that you have stated so clearly.
Nobody wants to make shitty applications on purpose and be there to see it's corpse rotting for eternity
Marcio - Thank you for sharing your experience with GitHub Copilot.
Cool article, thanks for sharing.
For me, personally the speed of typing code is not an issue, as a human, I can type slow compared with a bot, but quality is highly required, not speed 😅
I really liked your article! To add some additional alternatives to Github Copilot, Some open-source projects started using the remaining open-source training model, but most are stalled. Unless they were to receive funding, it would be pretty hard to achieve the same level of efficiency GitHub Copilot offers now. The latest version of its model is now available for licensing, preferably by Microsoft. As other companies can license OpenAI Codex, it’s possible we will have more alternatives in the future.
Tabnine aims to provide a similar product and already has a solid customer base (they claim to have more than 1 million developers using their product). They stand for using only open-source code with permissive licenses for their Open Source Trained AI model (MIT, Apache 2.0, BSD-2-Clause, BSD-3-Clause) from GitHub repositories or customer’s repositories/local code. You can also download the model to your machine and get even faster responses. Their Pro subscription comes in at $12 US per month.
In June 2022, Amazon launched a preview of CodeWhisperer, another tool similar to GitHub Copilot. This seems to be a legitimate competitor, but there’s not enough information to compare them so far.
It’s likely that we’ll have more competitors to evaluate in the future. For now, we’ll just say it’s “to be continued…”
I would like to share this article from my colleague Rafael Goulart, which presents a very interesting and interactive testing of GitHub Copilot in a small full-stack project using PHP and JavaScript: scalablepath.com/full-stack/ai-pai... He shared some real-life examples and some other pros and cons that may be useful if you are interested in these kind of tools.
If you let robot written code go into production without noticing it, it makes it kinda your fault.
Sure, I’m responsible for all my commits. That’s exactly my point.
Never said it wasn’t my fault. It’s totally my fault. I’m the one in control. So in control that I can choose not to use it anymore.
Cool, understood. Never meant to sound rude. Cheers!