How to Automate Software Engineering • Hacker News Board

Re: How to Automate Software Engineering

#51

Post by owebmaster » Sat, May 31, 2025, 9:38 PM UTC

belZaah wrote: post #42

Y’all need some Brooks in your lives. The No Silver Bullet essay explains very clearly, why the ability to string syntactically correct commands together does not equal the ability to write code.

  Artificial intelligence. Many people expect advances in artificial intelligence to provide the revolutionary breakthrough that will give order-of-magnitude gains in software productivity and quality. I do not. To see why, we must dissect what is meant by “artificial intelligence” and then see how it applies.

  Parnas has clarified the terminological chaos:
  Two quite different definitions of AI are in common use today. 
  AI-1: The use of computers to solve problems that previously could only be solved by applying human intelligence. 
  AI-2: The use of a specific set of programming techniques knows as heuristic or rule-based programming. In this approach human experts are studies to determine what heuristics or rules of thumb they use in solving problems. . . . The program is designed to solve a problem the way that humans seem to solve it.

  The first definition has a sliding meaning. . . . Something can fit the definition of AI-1 today but, once we see how the program works and understand the problem, we will not think of it as AI anymore. . . . Unfortunately I cannot identify a body of technology that is unique to this field. . . . Most of the work is problem-specific, and some abstraction or creativity is require to see how to transfer it

great read: https://worrydream.com/refs/Brooks_1986_-_No_Silver_Bullet.p...

Re: How to Automate Software Engineering

#52

Post by ege_erdil » Sun, Jun 01, 2025, 1:25 AM UTC

AnimalMuppet wrote: post #50

ege_erdil wrote: post #47

Earlier quoted context omitted.

what makes you sure about that?

At the risk of coming across as a smart aleck, 40 years of experience building software. What experienced software engineers have is a sense of taste - this looks like good code and/or design, that doesn't. But they don't have data ; they have, at best, a couple of anecdotes. It's more a sense of "that was harder to work with than it should have been; that approach seems to have drawbacks". But you only get a few exa…

that's not the relevant data i'm talking about

how much real-world data do you think went into the evolution of the human brain and all its learning algorithms?

having 40 years of experience building software gives you no more insight into that than having 40 years of experience using language gives you insight into where your language skills come from

Re: How to Automate Software Engineering

#53

Post by jongjong » Sun, Jun 01, 2025, 7:21 AM UTC

martin-t wrote: post #26

jongjong wrote: post #17

I think the conclusion about software engineering having the potential of being both the first and last job to be automated by AI makes sense. It is an extremely unequal profession. The skill difference between a junior with 1 year of experience and a senior with 20 years of both diverse and deep experience (of relentless striving) is massive. They have very different capabilities. The level of software sophisticatio…

Almost 15 years in and I feel like I am still just scratching the surface. But most of the lessons learned during that time go towards quality, not quantity or speed. The current trend with LLMs (and eventually maybe AI) seems to be doing what humans can do worse but significantly faster and cheaper. Unfortunately, not everyone needs or cares about safety, security or correctness. I am afraid software will mirror the…

This resonates. Quality of mass market products decrease to the upper limit of the average person's perception/expectations... Which itself seems to be declining over time.

Re: How to Automate Software Engineering

#54

Post by AnimalMuppet » Sun, Jun 01, 2025, 12:30 PM UTC

ege_erdil wrote: post #52

AnimalMuppet wrote: post #50

Earlier quoted context omitted.

At the risk of coming across as a smart aleck, 40 years of experience building software. What experienced software engineers have is a sense of taste - this looks like good code and/or design, that doesn't. But they don't have data ; they have, at best, a couple of anecdotes. It's more a sense of "that was harder to work with than it should have been; that approach seems to have drawbacks". But you only get a few exa…

that's not the relevant data i'm talking about how much real-world data do you think went into the evolution of the human brain and all its learning algorithms? having 40 years of experience building software gives you no more insight into that than having 40 years of experience using language gives you insight into where your language skills come from

The discussion is about algorithms automating writing software.

The human brain did not evolve to write software. It evolved to get a good enough answer fast enough to survive, not to solve the kind of problem that is involved in software.

Yeah, I know, the human brain is able to learn to do software. But that isn't what it evolved to do.