Conversation

Notices

1. Embed this notice
   clacke (clacke@libranet.de)'s status on Sunday, 05-Mar-2023 19:57:06 JST clacke
   > 1960s: "COBOL will let non-programmers make the software!"
   > 1980s: "4GLs will let non-programmers make the software!"
   > 2000s: "UML will let non-programmers make the software!"
   > 2020s: "AI will let non-programmers make the software!"
   farside.link/nitter/garybernha…
   • LinuxWalt (@lnxw48a1) {3EB165E0-5BB1-45D2-9E7D-93B31821F864} and GeniusMusing like this.
   • Embed this notice
     clacke (clacke@libranet.de)'s status on Sunday, 05-Mar-2023 22:05:06 JST clacke

     in reply to
     www.commitstrip.com/en/2016/08…
     GeniusMusing likes this.
   • Embed this notice
     badsynthesis (badsynthesis@infosec.exchange)'s status on Sunday, 05-Mar-2023 22:05:08 JST badsynthesis

     in reply to

     @clacke It's almost as if the problem isn't what they think.

     My view is that computers do what we say. Exactly. And only one thing at a time.

     Translating between natural language and that kind of domain is a job, regardless of language or tooling.

   • Embed this notice
     naugtur (naugtur@fosstodon.org)'s status on Sunday, 05-Mar-2023 22:49:28 JST naugtur

     in reply to

     @clacke
     2040: AI, could we please let programmers make the software?

     clacke likes this.
   • Embed this notice
     Sexy Moon (moon@shitposter.club)'s status on Sunday, 05-Mar-2023 22:49:29 JST Sexy Moon

     in reply to
     @clacke oh my god the replies, no one comes out looking good lol
     
     the inventor of uml comes in and says "I never said that"
     people reply with "what about Wix??? huh???"
     
     incidentally i had a manager in 1998 that got a program called java studio that let you drag and drop java "beans" in a gui to make programs without coding, she told me we wouldn't need programmers in a few years
     clacke likes this.
   • Embed this notice
     clacke (clacke@libranet.de)'s status on Sunday, 05-Mar-2023 22:50:40 JST clacke

     in reply to

     • Sexy Moon
     @Moon Maybe UML people didn't say it. There were definitely people who said it. ?
   • Embed this notice
     clacke (clacke@libranet.de)'s status on Wednesday, 08-Mar-2023 22:04:30 JST clacke

     in reply to

     • Jason_Dodd
     @Jason_Dodd We have gradually enabled software engineers to do more with less effort. Force multipliers. We don't flip switches to set machine code bits any more. We don't manually link libraries any more.
     
     No matter how much they are derided, no-code and low-code solutions have enabled non-programmers to improve their productivity and sometimes become programmers when necessary (more seldom full-blown software engineers, but that happens too and often leads to very well-rounded and innovative software engineers). Programmable spreadsheets are the prime example.
     
     Mostly though this has through a kind of Jevons' Paradox made human programmers even more sought for. Some tasks don't require them any more, so they have been made available for other tasks that do require expertise and an ability to translate requirements into implementation.
   • Embed this notice
     Jason_Dodd (jason_dodd@mastodon.social)'s status on Wednesday, 08-Mar-2023 22:04:31 JST Jason_Dodd

     in reply to

     @clacke This time it is true.

     Seems clear to me that what's different this types of iterations this time is that with every incremental improvement of AI more of the work that would require humans can be done by AI.

     We will run out of the next thing for humans to replace their current role with.

     Hope I'm gone before we find out if we can handle the transition.

   • Embed this notice
     clacke (clacke@libranet.de)'s status on Wednesday, 08-Mar-2023 22:04:32 JST clacke

     in reply to

     • nixCraft 🐧
     As we replace more and more software creation use cases with tooling, it's possible that we at some time reach some turning point where software developers would be less sought for, but I don't see it any time soon.
     
     ChatGTP and Copilot are not it. The unreliable output we have seen from ChatGTP and Copilot are not indicators that they are generic tools that can replace hand-crafted tools for specific programming scenarios. We can't leave them unsupervised any time soon. You'll need an expert to check their work, even if that expert might (or might not) be faster with the large language model assisting.
     
     Relevant comic (I made a repost with image caption, will find):
     ♲ mastodon.social/@nixCraft/1094…
   • Embed this notice
     mpjgregoire@mamot.fr's status on Wednesday, 08-Mar-2023 22:04:34 JST mpjgregoire

     in reply to

     @clacke The best way to get non-programmers making software is to train them in #Emacs .

     clacke likes this.
   • Embed this notice
     SlicerDicer (slicerdicer@bikeshed.party)'s status on Wednesday, 08-Mar-2023 22:04:38 JST SlicerDicer

     in reply to
     I really do not think its so much about the non programmer, as a programmer its very useful. It generates a lot of code that is very helpful.
     clacke likes this.
   • Embed this notice
     Rocket (rocket@shitposter.club)'s status on Wednesday, 08-Mar-2023 22:04:44 JST Rocket

     in reply to

     • Sexy Moon
     @Moon @clacke Idiots have said this about every new technology since BASIC. Now so-called "no code" solutions are being pushed everywhere. But this is what happens:
     
     1. Buy expensive no-code solution. Now business can make software without expensive programmers!
     2. This is hard to use, we need to hire specialists.
     3. Programmers are hired to use "no code" solution.
     4. Because they work for business, they are not subject to all the audit requirements that programmers are.
     5. "No code" programmers get stuff done faster because they don't have to document requirements, go through QA, have actual production access, don't have to talk to three departments to get anything done.
     6. Business is now tied to an expensive proprietary product.
     7. Auditors will eventually catch on. Auditors are really stupid though, so #5 could go on for a long time.
     clacke likes this.
   • Embed this notice
     clacke (clacke@libranet.de)'s status on Wednesday, 08-Mar-2023 22:04:45 JST clacke

     in reply to

     • Sexy Moon
     • Rocket
     @Rocket @Moon This is the core complaint of Luddites: Adding automation combined with removing the experts takes away expert jobs, sure, but it also makes for shoddy output and nobody is there to spot it.
   • Embed this notice
     badsynthesis (badsynthesis@infosec.exchange)'s status on Wednesday, 08-Mar-2023 22:06:14 JST badsynthesis

     in reply to

     @clacke I remember reading that one years ago, might have been a source of my current opinion :)

     clacke likes this.
   • Embed this notice
     clacke (clacke@libranet.de)'s status on Wednesday, 08-Mar-2023 22:07:32 JST clacke

     in reply to

     • 🌪 MikeMathia.com 📡
     > To replace programmers with Robots, clients will have to accurately describe what they want.
     > We're safe.
     farside.link/nitter/BUDESCODE/…
     
     /via ioc.exchange/@mikemathia/10996… @mikemathia
   • Embed this notice
     Cruiser (cruiser@expressional.social)'s status on Wednesday, 08-Mar-2023 22:07:34 JST Cruiser

     in reply to

     @clacke When ever the competent insight to understand, right level of abstraction for simplicity and robust design are omitted things become inadequate pieces of crap without ever enabling satisfaction or value for users

     But of course new full-stack engineers may get more diluted experiences from more trial and error with new tools matured for something else - the idea of a tech silver-bullet never goes away ?

     clacke likes this.
   • Embed this notice
     clacke (clacke@libranet.de)'s status on Wednesday, 08-Mar-2023 22:10:08 JST clacke

     in reply to

     • Jason_Dodd
     @Jason_Dodd I'm not seeing it. There's no qualitative difference, it's just more and less reliable automation.
     
     Maybe one day, but these increasingly complex autocompletes are not a path toward AGI.
   • Embed this notice
     Jason_Dodd (jason_dodd@mastodon.social)'s status on Wednesday, 08-Mar-2023 22:10:09 JST Jason_Dodd

     in reply to

     @clacke I've lived through it since 1989. But I do think AI adds a different twist in that the programmers that are needed will be increasingly AI and not human.

   • Embed this notice
     purp (purp@mastodon.social)'s status on Wednesday, 08-Mar-2023 22:10:11 JST purp

     in reply to

     @clacke you could add:

     > 1970s: "SQL will let non-programmers [make/not need] the software!"

     clacke likes this.
   • Embed this notice
     clacke (clacke@libranet.de)'s status on Wednesday, 08-Mar-2023 22:12:17 JST clacke

     in reply to

     • Jason_Dodd
     @Jason_Dodd I used to think so. I got over it. =)
     
     I've seen so many humans do a shoddy job because they just don't care and a LLM can't replace even them.
   • Embed this notice
     Jason_Dodd (jason_dodd@mastodon.social)'s status on Wednesday, 08-Mar-2023 22:12:18 JST Jason_Dodd

     in reply to

     @clacke I could be wrong, but I think we're nearing the tipping point where the AIs can do the part I've traditionally done.

   • Embed this notice
     Sandra (sandra@idiomdrottning.org)'s status on Wednesday, 08-Mar-2023 22:17:52 JST Sandra

     in reply to
     @clacke
     
     Arguably a lot of non-programmers did make software with COBOL
     clacke likes this.
   • Embed this notice
     chris (chris@s.the-brannons.com)'s status on Wednesday, 08-Mar-2023 22:18:43 JST chris

     in reply to

     • Sandra
     @Sandra @clacke
     So did BASIC and some forms of batch file.
     
     I used 4DOS batch language to write the first program I ever made for
     others to use. I was at a center for the blind. The program took a CD-ROM
     full of ebooks that were zip files each containing a single .txt file.
     Using a catalog, it organized that collection into directories named for
     authors, with the extracted .txt files renamed to contain the title. Yeah,
     the input was 8.3-style DOS filenames, and the output used Win 9x long
     filenames.
     
     I was doing most of this in my spare time, writing code on a PDA-like device
     with a Z80 microprocessor, no 4DOS interpreter in sight. Essentially I was
     coding by reading the 4DOS documentation.
     When I did have access to a PC with 4DOS, I'd test pieces of the thing.
     
     I had it make little beeps to indicate progress. Wanna guess how exciting
     it was to watch that thing run and hear those silly beeps?
     
     I was no programmer at the time; just a person with a brain and great reading
     comprehension.
     clacke likes this.
   • Embed this notice
     clacke (clacke@libranet.de)'s status on Wednesday, 08-Mar-2023 22:18:43 JST clacke

     in reply to

     • Sandra
     • chris
     @chris @Sandra Wow, could 4DOS handle long file names? Or was this running in 4NT?
   • Embed this notice
     chris (chris@s.the-brannons.com)'s status on Wednesday, 08-Mar-2023 22:19:29 JST chris

     in reply to

     • Sandra
     @Sandra @clacke
     All I remember is that it was 4DOS running under Win 95, and it handled
     long filenames just fine.
     This was a long time ago and a lot of the details are fuzzy. I wish
     I still had that code.
   • Embed this notice
     clacke (clacke@libranet.de)'s status on Wednesday, 08-Mar-2023 22:19:29 JST clacke

     in reply to

     • Sandra
     • chris
     @chris @Sandra Oh yeah, I remember now that some command-line tools under Windows 95 were actually capable of doing long filenames. So 4DOS could do that too, that's very cool.
   • Embed this notice
     Sandra (sandra@idiomdrottning.org)'s status on Wednesday, 08-Mar-2023 22:19:30 JST Sandra

     in reply to

     • chris
     @clacke
     
     CD-ROM makes me guess W95 w/ DOS 7. Elucidate us, @chris!