AI software development: Productivity revolution or fraught with risk? AI in software development has evolved rapidly since GitHub Copilot caught the world's attention with its June 2021 preview – and shows no sign of slowing down. At the same time, worries abound, and not only from developers whose jobs may or may not be under threat. Hallucinations, security issues, alleged copyright violations …
But I want refactoring! Bwaah! You promised me this would make my coding life easier! (Throws rattle out of pram).
Hey ho.
So this means I can ignore "AI coding" for a while longer, with a clean conscience, until they sort that out: at the moment, I have a morass of code filled with multiple (but very different) ways of "doing the same thing"[1] so even as I work on features the biggest, hugest, brobdingnagian part of the coding task is to refactor those.
Maybe if I'd spent my days generating web sites this "AI Coding Revolution" would be more useful?
[1] e.g. multiple lexers connected to multiple input streams - so some can do an "include" at the simplest, lowest, level and some can not do it themselves at all. Multiple "little languages" built on those that have their own concept of where to store variables, so passing context from one to the other is a PITA. And all of this marked up with comments saying "this is being done in a few rush to get XYZ out of the way, will fix Real Soon Now": there are no more XYZs and "Soon" has finally come, after, let us see, oops - a decade and a bit!
Data-center capacity WILL overshoot. We've seen this movie many times.
Gaslighting on an enormous scale WILL continue and prices across all of IT WILL increase to make up for the colossal mistakes.
Useless AI tat-for-the-masses WILL be forced on us to keep stock prices up.
We've seen this movie many, many times. Huge expensive mistakes will be made and us peons will pay the price, just like we always do.
I'm a new coder but I love using AI. So far Amazon Q seems really good, until it's not. Does seem to hallucinate regularly and sometimes takes a convoluted approach instead of the simplest. I find I am honing my use of it and actually using it less but where it saves most time - such as being stuck and not having to spend so much time pouring over docs and examples as it pulls them out for me. But I agree you have to look carefully at what it produces, it's regularly a bit off but a good pointer towards what's needed. I guess if you have 10 years experience in a language then it will rarely help. I really like its ability to look at a block of code that isn't doing what's expected, especially if involving API calls, then respond with what is wrong and suggest a fix, sure doesn't always work but it kind of leads you to find the problem.
