No, they have a place where stochastic algorithms are necessary. For writing a hello world application, no stochastic algorithm is necessary. Comparing compilers with LLMs is comparing apples with oranges.
I think it will become more apparent over time. But consider that the practice of software engineering is a stochastic process. Give 10 different engineers the same goal, and you’ll get 10 different solutions.
You’re saying “we have to draw the line”. If I’m understanding the discussion at hand, I’m saying: we don’t. But I’d like to clarify what line it is you think we need to draw.
I think this is an interesting discussion to have, but if it’s not enjoyable to you, we can end it here. Cheers.
Edit: reading back again, I think you’re saying we need to draw the line and only use stochastic solutions for problems that necessitate them. That’s fair, ex. it’s inefficient, and error prone to invoke an AI to sort a list.
But rarely do humans have unsorted, well tabulated lists that they need sorted. Most people’s goals are stochastic. They have photos that need organized by location, event, content, etc. They have hundreds of emails from customers all asking the same trivial questions in different ways. They are going to meet a friend at the store across town and need to give an ETA.
Goals are only well defined if you only operate inside the well-defined space of formal languages. But the formal languages don’t exist for their own sake, at the end of the day, we built computers to solve amorphous, difficult to describe, human problems, and the messiness of software engineering has always reflected that.
If a SW dev applicant gives a 20-file generated output for a 20-line assessment problem and can’t explain single lines of “their” code, either what they should be doing or why “they” had written it, it’s gonna be a no from me, dawg. A standard problem might have different solutions, but fixing the issue of the day to the satisfaction of a rabidly vocal customer base might have one at most, and it will change multiple times on a whim.
So the LLM might have helped them cheat their way to an MSc, but there’s no cheating your way through real life.
No, they have a place where stochastic algorithms are necessary. For writing a hello world application, no stochastic algorithm is necessary. Comparing compilers with LLMs is comparing apples with oranges.
I think it will become more apparent over time. But consider that the practice of software engineering is a stochastic process. Give 10 different engineers the same goal, and you’ll get 10 different solutions.
At that rate me walking to the store is stochastic because a grand piano could fall on my head. We have to draw the line at some logical point.
What line are we trying to draw exactly? I think that’s the part I’m still confused on.
Yes, walking to the store is a stochastic process. Ask anyone working at google maps.
I think you’re arguing for arguing’s sake. If you don’t see the point by now, then I am unable to make it more clear. I’m sorry.
You’re saying “we have to draw the line”. If I’m understanding the discussion at hand, I’m saying: we don’t. But I’d like to clarify what line it is you think we need to draw.
I think this is an interesting discussion to have, but if it’s not enjoyable to you, we can end it here. Cheers.
Edit: reading back again, I think you’re saying we need to draw the line and only use stochastic solutions for problems that necessitate them. That’s fair, ex. it’s inefficient, and error prone to invoke an AI to sort a list.
But rarely do humans have unsorted, well tabulated lists that they need sorted. Most people’s goals are stochastic. They have photos that need organized by location, event, content, etc. They have hundreds of emails from customers all asking the same trivial questions in different ways. They are going to meet a friend at the store across town and need to give an ETA.
Goals are only well defined if you only operate inside the well-defined space of formal languages. But the formal languages don’t exist for their own sake, at the end of the day, we built computers to solve amorphous, difficult to describe, human problems, and the messiness of software engineering has always reflected that.
If a SW dev applicant gives a 20-file generated output for a 20-line assessment problem and can’t explain single lines of “their” code, either what they should be doing or why “they” had written it, it’s gonna be a no from me, dawg. A standard problem might have different solutions, but fixing the issue of the day to the satisfaction of a rabidly vocal customer base might have one at most, and it will change multiple times on a whim.
So the LLM might have helped them cheat their way to an MSc, but there’s no cheating your way through real life.
What you’re describing is “vibe coding”. For the difference between vibe coding and engineering, I’ll refer you to my previous comment.
Edit: had problems with opening your link, replying to your linked comment now…