Comment Re:What about 'new' stuff (Score 1) 116
A programmer faced with a new language, OS, API or whatever has to sit down and learn it from documents, not existing examples. Without programmers programmers creating stuff with the new thing there is nothing for the AIs to be trained on.
This is a good insight, and one I've been thinking about lately.
I have a small development team (two senior developers), and have been tasked by our president to research the potential use cases of AI in our organization. My developers, who don't really need AI help, have been skeptical about trying out code assistants. As a former (15 years ago) developer, I've been using Copilot to help me brush up on what's changed since I've been away (answer: a lot, aside from the fundamentals).
My takeaways after a few weeks of experimentation:
* AI is not going to replace senior developers / architects anytime soon. As ukoda said above, LLMs are only going to regurgitate, and we still need humans for novel solutions.
* Even senior developers should be using coding assistants to accelerate the "boring stuff". I told my team that they should be letting the AI do the things they knew how to do 10 years ago, so they can spend their time doing the stuff that they can still do better than the AI.
* Junior developers are going to be caught up in a chicken-and-egg cycle. From the time they start out in school, novice programmers are going to let the AI generate code that they don't understand, and it's going to end up in production. A lot of it will work well enough that management is going to reduce/eliminate junior positions. Without the ability to gain experience, there won't be enough seasoned devs available to properly examine the AI code output for security issues, bugs.
* If I'm correct about my previous point, manual programming skills may become an expensive niche similar to COBOL in the 21st century. It'll be a great skill to have, because you won't have a lot of competition.
