"One of the biggest problems with using LLMs as a google replacement for programming,
is that getting zero relevant results on google used to be a signal that you had the wrong idea about the root cause. Where as LLMs will happily indulge any terrible idea you suggest." -Nic Barker
it's the concept of "Gell-Mann amnesia". You realize they are wrong whenever they talk about stuff you understand, and yet you still believe them when they talk about stuff you don't know about.
nailed it with the "tedious redundant boilerplate" comment
AI has saved me so much drudgery reading docs and fruitlessly googling stuff
but when I'm actually programming I find it's of limited use. Sometimes I'll start a project in AI, but it won't really get moving until I replace most of the AI generated stuff.
Anyone that's tried to use AI for something that doesn't already have a stack overflow post for will know its flaws. I think AI is actually useful for understanding poorly written documentation though
We spent a lot of years writing books about using laziness as some sort of architectural superpower. "I'm lazy, so I'm going to abstract everything so I never have to repeat anything and use two-letter variable names."
"Chatbots are like any other machine - they're either a benefit or a hazard. If they're a benefit, it's not my problem." ~some guy that ran around chasing chatbots, supposedly
These zealot supporters of AI are so cultish its sad. They're basically pumping the propaganda rather than what AI can actually do. They're willing spokespersons for big billionaire chumps.
I'm a beginner, programming as a hobby, LLMs are a double edged sword. It can get you an answer quickly that might solve the problem you have by coincidence but your actual problem is somewhere else. So you either blow through insane amounts of tokens to give it full context in hopes that it catches your problems, or you try and solve it yourself. When I started, trying to figure out how to put a window on the screen was insane. Something I thought would be easy is a rabbit hole I'm still not sure what is going on with it. LLMs at least pointed me in a direction. The problem ultimately is it gives you the answer without the work. I learn more and have it stick (even in my late 30's) when I work through it myself, and try to reason out stuff. I've come to the opposite conclusion. I think it is the people who know what they're doing that get more use out of it if they use it, but at that point you just do it yourself. I do agree though, from the outside looking in, it seems like it has slowed down, and the improvements seem to be people constraining them enough to keep them focused, more so than the models doing the heavy lifting.