|
It might turn out that way, but it applies just as much between a manager who needn't fear ever having to work for the person being chewed out. The best boss I ever had put it as, "So hard to build [trust], so easily destroyed."
|
|
|
|
|
No doubt, it definitely feels good! That's why I also use this tactic to motivate employees.
Think about it 
|
|
|
|
|
I mean, I don't think it was done with the intent to motivate me, as I'm pretty motivated to begin with - goes with loving the job. I think I just have a great client. 
Check out my IoT graphics library here:
https://honeythecodewitch.com/gfx
And my IoT UI/User Experience library here:
https://honeythecodewitch.com/uix
|
|
|
|
|
Reminds me of when a guy in the support department for our major development environment at that time sent me a Christmas greeting to his favorite customer: The bug reports from us (for all practical purposes: me) were exemplary: Clearly stated, reduced to the minimum required to reproduce the problem, and with lucid descriptions of how to progress to provoke the bug. If all customers had been like that, he would have been a happy man.
That is the single Christmas greeting I don't think I will ever forget.
Religious freedom is the freedom to say that two plus two make five.
|
|
|
|
|
Part of my job at the tiny company I work for is to look after the licenses we sell our customers and to help them get going, or upgrade to newer versions, that sort of thing, and I can absolutely do some of the more advanced tech support for any part of the code that I've actually written (and some I haven't).
Throughout the years some customers have gone out of their way to say how impressed they are with the support they're getting, our dedication to quickly finding solutions or going right ahead and implementing a bug fix that gets released within hours of a problem being reported. Those always make my day.
I never use a condescending tone, or treat customers like they're idiots, and always try to walk them through solutions just as if I was sitting right next to them at their desk and having a conversation.
|
|
|
|
|
Have you tested it? Any conclusions?
Quote: Did you know that research shows developers who use GitHub Copilot complete tasks 2X faster compared to benchmarks? Plus, the mental energy they save helps them feel more satisfied and have more fun with their work.
That's from an email I received from GitHub (MS)...
To be honest I can't say anythin good about Copilot - all 5 times I asked a question it bomarded me with total nonsence (and not working, and mostly not even compiling) code solutions, meanwhile not paying attention to the details but hanging on the recognized keywords (like picking 'working' from 'not working')...
However, I will form a firm oppinion after at least 50 times of using it...
"It never ceases to amaze me that a spacecraft launched in 1977 can be fixed remotely from Earth." ― Brian Cox
|
|
|
|
|
I may just be old and set in my ways, but I don't trust it (or any "AI") to write code, and even if I could, I doubt it would write code the way I write code. I don't like it, and I have little use for it. At best, it's a fancy google machine but stackoverflow, github and codeproject can typically answer my questions, and I trust them more.
Check out my IoT graphics library here:
https://honeythecodewitch.com/gfx
And my IoT UI/User Experience library here:
https://honeythecodewitch.com/uix
|
|
|
|
|
I was thinking that it may shorten search for documents/ideas as it should have scande a wast amount of code...
So it is either bad at finding things or the avarage of the code on the web is very low 
"It never ceases to amaze me that a spacecraft launched in 1977 can be fixed remotely from Earth." ― Brian Cox
|
|
|
|
|
it's just a tool.
I would never release code without reviewing it first and/or adapting it.
CI/CD = Continuous Impediment/Continuous Despair
|
|
|
|
|
It's a tool that's statistically wrong more often than not as it pertains to coding questions.
If any of my other tools had that kind of failure rate I'd throw them away.
Check out my IoT graphics library here:
https://honeythecodewitch.com/gfx
And my IoT UI/User Experience library here:
https://honeythecodewitch.com/uix
|
|
|
|
|
We don't use it at work because of the potential liability of using copyrighted code. I understand that MS will not indemnify anyone against lawsuits for copyright violation if Copilot outputs a block of copyrighted code.
Even if this were not an issue, at the current state of the art I have to thoroughly review any suggestions by Copilot and its ilk. By the time I've done that, I might as well have written the code by myself.
Copilot will only be useful if and when it can reliably output entire blocks of code, rather than snippets. At which time, see my first point.
Freedom is the freedom to say that two plus two make four. If that is granted, all else follows.
-- 6079 Smith W.
|
|
|
|
|
I didn't actually meant to let it write code - more of a fast and focused search... not impressed there...
"It never ceases to amaze me that a spacecraft launched in 1977 can be fixed remotely from Earth." ― Brian Cox
|
|
|
|
|
Among other things - and this is more a general AI concern I have - I'm concerned about model collapse in AI as it is being trained with online content ... generated with AI.
Nature had an article about it recently. I call it AI incest.
As code online that is generated with tools like ChatGPT and Copilot becomes more prolific what happens to the models?
This isn't my biggest issue with it, but it does seem that AI is destined to eat itself.
Check out my IoT graphics library here:
https://honeythecodewitch.com/gfx
And my IoT UI/User Experience library here:
https://honeythecodewitch.com/uix
|
|
|
|
|
Interesting point. Could it be that the only reason for humans' long-term existence is to "salt" the AI models?
Freedom is the freedom to say that two plus two make four. If that is granted, all else follows.
-- 6079 Smith W.
|
|
|
|
|
I don't think I understand the question? The way I'm interpreting it doesn't make sense, but I can't figure it out. The way I'm reading it suggests that AI has been around for the whole of human history  but I know that's not what you're saying.
Check out my IoT graphics library here:
https://honeythecodewitch.com/gfx
And my IoT UI/User Experience library here:
https://honeythecodewitch.com/uix
|
|
|
|
|
Humans evolved, then they built AIs. The AIs eventually take over, and keep humans only so they can add some randomness into the AIs' models.
Isaac Asimov wrote a good short story on a similar theme, The Last Answer. It's not as well-known as his The Last Question, but worth reading IMO.
Freedom is the freedom to say that two plus two make four. If that is granted, all else follows.
-- 6079 Smith W.
|
|
|
|
|
Daniel Pfeffer wrote: Even if this were not an issue, at the current state of the art I have to thoroughly review any suggestions by Copilot and its ilk. By the time I've done that, I might as well have written the code by myself.
This is my POV on "self driving" vehicles -- if I have to have my hands on the wheel and be paying attention 100% of the time, why do I need it?
|
|
|
|
|
I've been using it for the last couple of months. You have to treat it like you're an expert QA abuser here.
Context is EVERYTHING. The more detail, the better. If you don't describe in nauseating detail what you're working on, the app type, what you want the app to do, the problems you're having, ... It's going to do what we do when we come across an idiot in QA who doesn't know how to ask a question. It's going to guess at what you're doing as best it can.
But what you said about code that doesn't compile is true. If you're not watching what it's doing, it will spit out code that doesn't compile. I've had multiple occasions where it generated WPF XAML and used properties on classes that didn't exist. In these cases, I told it what it did wrong, and to give it credit, it did come up with corrected solutions, even ones that didn't involve just changing a property, but instead, rewrote the entire block using an entirely different method to solve the original problem.
It works best if you don't try and treat it like you're in QA. Treat it like it's a friend you tell everything to about your code and the results get better.
It's a tool like any other, where you have to learn to use correctly.
Would I trust it to write production quality code? Not a chance, but it can be a decent helper. We're in no danger at all of AI replacing us, or even replacing QA here.
|
|
|
|
|
I agree. When working on code with Copilot I start out with the code context. Then I explain my problem and include the compiler error message. Copilot commiserates and rewrite my code, improving it. Then it offers alternatives. The more descriptive your prompt, the better the answer. I never ever expect Copilot to design or architect for me. I can already barely remember the time when I had to look up obscure syntax and battle, in my case, Microsoft regarding documentation errors and get rebuffed with "Need more information" and "Not important enough" and issue is "closed".
|
|
|
|
|
I've found it to save some time with autocomplete when you have an existing example in the code to work from and it's a repetitive task.
E.g.
public void IsNumberEven(int number)
{
if (number == 0) return true;
if number == 1) return false;
if number == 2) return true;
// copilot will do well with autocomplete here
}
:P
|
|
|
|
|
I am thinking it could save some typing by using recursion and adding or subtracting 2 on each call!
Start posting multiple instances of cr@p like that on public repos and see how long it takes to suggest that…
|
|
|
|
|
I haven't used it, but I should and kina want to. AI is going nowhere. Your options are embrace it or become a dinosaur.
Jeremy Falcon
|
|
|
|
|
And another study I saw the other day contradicts that email: Study shock! AI hinders productivity and makes working worse • The Register
That study mirrors my experience, tbh. I've found "ai-powered" code tools to create more work than they take away because (a) I'd say about 50% of suggestions are completely wrong and the rest are either trivial or subtly wrong.
------------------------------------------------
If you say that getting the money
is the most important thing
You will spend your life
completely wasting your time
You will be doing things
you don't like doing
In order to go on living
That is, to go on doing things
you don't like doing
Which is stupid.
|
|
|
|
|
I can't say anything about Copilot, but I can about its Gemini counterpart, which is currently in the free stage.
The autocomplete works better than the standard autocomplete. But as always, read carefully - the bot doesn't understand the purpose of this piece of code. For repetitive tasks, it works great. But for longer functions, I need a lot of time to review.
Sometimes I was surprised. The function was written in a way I would never use. But the result was correct and sometimes even better.
I see these bots more in the area of learning than productive creation. But yes - they are fun too.
When it comes to learning other people's code, they are really a help.
|
|
|
|
|
CoPilot allowed me to have somewhat in-depth conversations about nuclear physics and criticality that Microsoft newest AI would not handle. Except for that, so far, all the AI either refuse to discuss controversial subjects or adhere to the official line. I find this problematic.
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Praise 
Rant 
Admin 
Submit an article or tip
