My employer Red Hat is one of many dabbling in AI code generation, and although I'm not directly involved in it, at least from what I've seen for the foreseeable future, I anticipate AI to be a helper and a supplement to "real" programmers and sysadmins. The AIs already do a fantastic job generating whole working functions for known solved problems, but they're nowhere (yet) near building whole systems from scratch or synthesizing solutions to brand new ideas. Want to iteratively calculate pi using the Gregory-Leibniz series? Easy. Literally just ask OpenAI in your language of choice, and it'll spit out working code today and demonstrate its output. I did exactly this a few days ago for fun. But what it you want a function to calculate some new theorem or pattern or behavior that isn't already exhaustively covered in textbooks and blogs and StackOverflow posts? Or using a whole new programming language or library? Good luck. As long as we're coming up with new ideas, we're going to need humans to program them. There's also a lot of context and integration work that is currently hard for AIs to understand and do well.
Remember: the AIs we're talking about today are still "just" big pattern recreation and re-synthesis engines. That goes for text, audio, images, etc. Yes, they're highly sophisticated and complex engines, but at the heart, that's all they are. Truly new concepts do not exist to them and need to be trained back into the models (and reinforced) over time.
I currently see AI code generation for programmers to serve a role a lot like Lightroom and Photoshop do for photographers and artists. Gone are the days where you had to physically treat and develop film in a darkroom or physically airbrush details. Sure, you could still do it by hand, and to learn the fundamentals maybe you should and there will always be niche hobbyists who choose to do it the old fashioned way, but both amateur and professionals alike will learn to use these new tools to experiment and iterate quickly, draw new inspirations, and ultimately produce better work. Yes, some people in tech will be displaced. There are plenty of people in IT and dev who are really just hiding out in a cubicle (virtually or literally) to collect a paycheck, basically relying on Google to get their jobs done. Their jobs, or a large part of them, will be among the first to go. The actual creators and innovators and system integrators likely still have a much longer career ahead of them.
|