Ask HN: What are your honest thoughts on AI tools replacing programmers
9 comments
I think the notion of what a programmer is will change. Programmers used to write punch cards, and then they wrote machine code, then code that had to be compiled, then interpreted code, and so on. It may just be that programmers will be paid to wrangle AI, or it may be that you use your skills to build tools out of AI agents. That's what excites me at the moment is building things that use AI agents to automate manual processes, and selling the service to others. This is more interesting than working for a big company as a software engineer.
historical observation: Key punch 'wrote' punch cards. Compilation was a 'simpify for the human how to describe it to a machine'. Interpreted was 'simplify how to describe it for the non-programmer without an engineering degree'.
AI can skip the 'intermediate process' of description -> programming language -> "machine binary". Still has to be 'trained'/'feed' appropriate info to be able to generate reasonable associations/inferences to do 'AI magic'.
Forseeable future, human will still need to understand & to describe the manual process to the AI agents. (vs. "let the programmer figure it out")
AI can skip the 'intermediate process' of description -> programming language -> "machine binary". Still has to be 'trained'/'feed' appropriate info to be able to generate reasonable associations/inferences to do 'AI magic'.
Forseeable future, human will still need to understand & to describe the manual process to the AI agents. (vs. "let the programmer figure it out")
> i find myself building some new niche tool almost daily. so the notion of ai "replacing programmers" seems kinda odd because most of my friends and I just still build stuff and jump in to tweak code here and there. our desire to build stuff and be on our IDEs every day still hasnt gone anywhere.
From scratch, try something in range of 1 million or more lines of code for basic functionality. aka OS, database management system, video editor, etc.
From scratch, try something in range of 1 million or more lines of code for basic functionality. aka OS, database management system, video editor, etc.
Why do people keep asking this? This question pops up on here almost everyday and the answers remain the same.
I’m guessing it’s a high-frequency trading sort of situation, trying to predict when a phase shift occurs in either the acceptance or rejection of AI coding by continuous pulsing of the collective consciousness of HN. Then correlating that with events such as the roll-out of AI releases, general hype, etc. to see what made a difference to cause a bit-flip. I think the questions are too ill-specified to gain much insight from, but asking is free.
Or “they” have so little general visibility into what’s actually happening at a grass roots level, that “they” are desperate for any news from the front. Again, asking is free. Let’s hope they don’t start using other HFT techniques.
Or “they” have so little general visibility into what’s actually happening at a grass roots level, that “they” are desperate for any news from the front. Again, asking is free. Let’s hope they don’t start using other HFT techniques.
For same reason AI controled surgery isn't going to replace surgical school experience.
Have to have backround to understand tool(s) using in order to supervise/know what's the right/appropriate tool for job(s)/task(s).
Have to have backround to understand tool(s) using in order to supervise/know what's the right/appropriate tool for job(s)/task(s).
Desperate to outsource their thinking, too lazy to punch a few keys and jostle a few brain cells into contacting one another.
its not that deep
god damn yall are mean lol
i find myself building some new niche tool almost daily. so the notion of ai "replacing programmers" seems kinda odd because most of my friends and I just still build stuff and jump in to tweak code here and there. our desire to build stuff and be on our IDEs every day still hasnt gone anywhere.
do you guys think we'll truly get to 0 manual coding or software engineering will slowly get reduced to infra/devops and high level architecture?