4 ms·
Even if AI gets to the point where I can lazily describe any software project I want and it will make it for me, even better than I described it, I'm 100% sure
by cwalv 2y ago
Even if AI gets to the point where I can lazily describe any software project I want and it will make it for me, even better than I described it, I'm 100% sure that nothing will stop me from moving on to other projects short of death/being incapacitated.
I doubt AI will be getting to that point in my lifetime, but if it did, it would be amazing
- shortrounddev2 2y agoIf AI got to that point in my lifetime I'd quit software engineering. That sounds like a fucking mind numbing job. I'd go be a goat farmer or something before I become a professional chat bot chatter
- cwalv 2y agoMost of software engineering (differentiating from programming) is just iterating on a model of a problem space in enough detail that computers can be applied to automate a lot of the drudgery. Software is nowhere close to done "eating the world". I see endless opportunities to apply software to remove drudgery; anything that helps me do that faster is a win.
- shortrounddev2 2y agoEveryone has a level of abstraction they're comfortable with. I enjoy working in C++, and I tolerate working in Javascript for work. I wouldn't be good at writing assembly, and I would dread talking to chatbots to tell them what code to write. I enjoy solving problems, not describing them to chatbots.
- cwalv 2y agoIMO it's hardly more of a stretch to say that your C++ compiler is just an obtuse chatbot. If it doesn't complain with template error vomit, you understand its response is 'done'. Otherwise, it's chatting back the errors of your ways.
- shortrounddev2 2y agoI can model the execution of a C++ program as it relates to the underlying stack, heap, CPU, etc. in my mind as I'm coding it. I cannot with whatever LLMs do. Anyways, it's just not real engineering to do so. A world in which we spend all day talking to chatbots is simply not a world I want to be a part of
- cwalv 2y ago> I can model the execution of a C++ program as it relates to the underlying stack, heap, CPU, etc. in my mind Fair enough. It's a very approximate model tho; even C++ is operating at a very high level of abstraction. To me, what's more important is what it does, not how it does it. To use an LLM effectively you absolutely have to understand what it's doing; for software, that's the code it generates and the domain/ecosystem the software operates in. Conceptually, it's not that different from what senior ICs have been doing for ages, when they delegate to less-senior ICs