[COFF] Re: The era of general purpose computing (Re: AIX moved into maintainance mode

[COFF] Re: The era of general purpose computing (Re: AIX moved into maintainance mode

[segaloco via COFF]

COFF'd

I wonder if we'll see events around 2038 that renew interest in conventional computing.  There are going to be more public eyes on vintage computers and aging computational infrastructure the closer we get to that date methinks, if even just in the form of Ric Romero-esque curiosity pieces.

Hopefully the cohort of folks that dive into Fortran and Cobol for the first time to pick up some of the slack on bringing 2038-averse software and systems forward will continue to explore around the margins of their newfound skills.  I know starting in assembly and C influenced me to then come to understand the bigger picture in which those languages and their paradigms developed, so hopefully the same is true of a general programming community finding itself Fortran-and-Cobol-ish for a time.

- Matt G.

------- Original Message -------
On Saturday, January 21st, 2023 at 10:43 AM, Luther Johnson <luther@makerlisp.com> wrote:


> Yes, I know, but some of that SW development is being automated ... I'm
> not saying it will totally go away, but the numbers will become smaller,
> and the number of people who know how to do it will become smaller, and
> the quality will continue to deteriorate. The number of people who can
> detect quality problems before the failures they cause, will also get
> smaller. Not extinct, but endangered, and we are all endangered by the
> quality problems.
> 
> On 01/21/2023 11:12 AM, arnold@skeeve.com wrote:
> 
> > Real computers with keyboards etc won't go away; think about
> > all those servers running the backends of the apps and the
> > databases for the cool stuff on the phones. Someone is still
> > going to have to write those bits.
> > 
> > Arnold
> > 
> > Luther Johnson luther@makerlisp.com wrote:
> > 
> > > Well, that's a comforting thought, I hope it goes that way.
> > > 
> > > On 01/19/2023 06:10 PM, John Cowan wrote:
> > > 
> > > > On Thu, Jan 19, 2023 at 5:23 PM Luther Johnson <luther@makerlisp.com
> > > > mailto:luther@makerlisp.com> wrote:
> > > > 
> > > > Computers that are not smart phone-like are definitely on the
> > > > endangered
> > > > species list. You know, the kind on a desk, with a keyboard ...
> > > > 
> > > > I don't have statistics for this, but I doubt it. Consider amateur
> > > > radio, which has been around for a century now. Amateur stations are
> > > > an ever-shrinking fraction of all transmitters, to say nothing of
> > > > receivers, but in absolute terms there are now more than 2 million
> > > > hams in the world, which is almost certainly more than ever.

[COFF] Re: [TUHS] The era of general purpose computing (Re: AIX moved into maintainance mode (COFFed)

[David Arnold]

(Move to COFF)

> On 22 Jan 2023, at 05:43, Luther Johnson <luther@makerlisp.com> wrote:
> 
> Yes, I know, but some of that SW development is being automated ... I'm not saying it will totally go away, but the numbers will become smaller, and the number of people who know how to do it will become smaller, and the quality will continue to deteriorate. The number of people who can detect quality problems before the failures they cause, will also get smaller. Not extinct, but endangered, and we are all endangered by the quality problems.

Is it possible that this concern mirrors that of skilled programmers seeing the introduction of high-level languages?

I’ve played with ChatGPT, and the first 10 minutes were a bit confronting.  But the remainder of the hour or so I played overcame my initial concerns.

It’s amazing what can be done, especially with Javascript or Python, when you ask for something that’s fairly simple to define, and in a common application area.  You can get reasonable code, refine it, ask for an altered approach, etc.  It’s probably quicker than writing it yourself, especially if you’re not intimately familiar with the library being used (or even the language).

But … it pretty quickly becomes clear that there’s no semantic understanding of what’s being done behind it.  And unless you specify what you want in pretty minute detail, the output is unlikely to be what you want.  And, as always, going from a roughed-out implementation of the core functionality to a production-ready program is a lot of work.

In the end, it’s like having an intern with a bit of experience, Stack Overflow, and a decent aptitude driving the keyboard: you still have to break down the spec into small, detailed chunks, and while sometimes they come back with the right thing, more often, you need to walk through it line by line to correct the little mistakes.

I’m looking forward to seeing a generative model combined with static analysis, incremental compilation, unit test output: I think it will be possible for a good programmer to multiply their productivity by a few times (maybe even 10x, but I don’t see 100x).  There’ll still be times when it’s simpler to just write the code yourself, rather than trying to rephrase the request.

All of which makes me think of the assembly language to high-level language transition ...




d

[COFF] Re: The era of general purpose computing (Re: AIX moved into maintainance mode (COFFed)

[Stuff Received]

On 2023-01-22 00:23, David Arnold wrote (in part):
 > I’ve played with ChatGPT, and the first 10 minutes were a bit
 > confronting.  But the remainder of the hour or so I played
 > overcame my initial concerns.
 >
 > It’s amazing what can be done, especially with Javascript or
 > Python, when you ask for something that’s fairly simple to
 > define, and in a common application area.  You can get reasonable
 > code, refine it, ask for an altered approach, etc.  It’s probably
 > quicker than writing it yourself, especially if you’re not
 > intimately familiar with the library being used (or even the
 > language).
 >
 > But … it pretty quickly becomes clear that there’s no semantic
 > understanding of what’s being done behind it.  And unless you
 > specify what you want in pretty minute detail, the output is
 > unlikely to be what you want.

The comment by Derek Jones 
(https://shape-of-code.com/2023/01/15/coding-mistakes-made-by-chatgtp/) 
may be of interest here.

N.

[COFF] Re: The era of general purpose computing (Re: AIX moved into maintainance mode (COFFed)

[segaloco via COFF]

One of the higher ups at work has suggested quite an interest in ChatGPT as a code documentarian.  I'm wary to let that thing near my code bases, but I'm a stickler for style as much as function, so a bit you could also tell to make pretty, readable code would be interesting.

I see most effective applications in documentation and cleanup, as a bot could probably be made to spit out pretty consistent documentation of functions to the tune of "always describe an if else option with this verbiage" or "add a mention of the stack depth added by every routine", and then cleanup like enforcing/removing braces on one liners, resorting procedures to group more relevant operations (when not order-dependent), and perhaps better intelligent editors that don't need as much configuration to smartly support many programming languages.  I'd love an editor that allows for rich editing of assembly alongside C.

Hopefully the direction the improvements on productivity go are "Gee my workers can work so efficiently now they deserve the time this frees up for them." although my fear is "Gee my workers can work so efficiently now I only need half of them."  I only support automation that makes workers' lives easier, not automation that makes executives bottom lines better at the cost of jobs.

- Matt G.

------- Original Message -------
On Sunday, January 22nd, 2023 at 7:40 AM, Stuff Received <stuff@riddermarkfarm.ca> wrote:


> On 2023-01-22 00:23, David Arnold wrote (in part):
> 
> > I’ve played with ChatGPT, and the first 10 minutes were a bit
> 
> > confronting. But the remainder of the hour or so I played
> 
> > overcame my initial concerns.
> 
> > It’s amazing what can be done, especially with Javascript or
> 
> > Python, when you ask for something that’s fairly simple to
> 
> > define, and in a common application area. You can get reasonable
> 
> > code, refine it, ask for an altered approach, etc. It’s probably
> 
> > quicker than writing it yourself, especially if you’re not
> 
> > intimately familiar with the library being used (or even the
> 
> > language).
> 
> > But … it pretty quickly becomes clear that there’s no semantic
> 
> > understanding of what’s being done behind it. And unless you
> 
> > specify what you want in pretty minute detail, the output is
> 
> > unlikely to be what you want.
> 
> 
> The comment by Derek Jones
> (https://shape-of-code.com/2023/01/15/coding-mistakes-made-by-chatgtp/)
> may be of interest here.
> 
> N.