From: Mark Knecht <markknecht@gmail.com>
To: gentoo-user@lists.gentoo.org
Subject: Re: [gentoo-user] Books about making shell scripts and other nifty commands.
Date: Thu, 5 Jun 2025 09:16:50 -0700 [thread overview]
Message-ID: <CAK2H+ed88JaNrrHgZt5Bpw=y7zaucsmktgmgbJwGTGF0DAkm=Q@mail.gmail.com> (raw)
In-Reply-To: <d508c400-d497-447f-99e8-452314222dfd@gentoo.org>
On Thu, Jun 5, 2025 at 9:03 AM Eli Schwartz <eschwartz@gentoo.org> wrote:
>
> On 6/5/25 11:32 AM, Mark Knecht wrote:
> > A book is one way to go, especially if someone trusted recommends
> > the title.
> >
> > As for me - and this isn't bash scripting - but I've learned Python just
> > using videos on YouTube coupled with asking AI Helpers {Chat GTP,
> > Gemini, Claude, Deepseek, Copilot} either questions I want answered
> > or, more importantly, providing it with some code I wrote and asking it
> > two basic questions:
> >
> > 1) What does this code do?
> >
> > 2) Are there any problems this code might cause or side effects I
> > should be aware of?
> >
> > Those two questions plus a copy of my real data I want to work on
> > so that it doesn't make a mistake and delete or corrupt my real data
> > have allowed me to make a lot of progress without spending a
> > penny. I suspect this would work well for you if it suits your learning
> > style.
> >
> > I see plenty of videos on YouTube about bash scripting.
> >
> > All of the AI helpers will answer 20-30 questions a day for free so
> > there's no cost involved, and you can ask each of them the same
> > questions to get multiple points of view on solving your problems.
>
>
> Artificial non-intelligence is based on a sophisticated
> https://en.wikipedia.org/wiki/Mad_Libs generator. It is a program
> designed to produce pseudorandom output streams with a weight towards
> grammatically correct output.
>
> It possesses no knowledge or reasoning and literally *cannot* possess
> the attribute of being correct. It is very good at superficially looking
> correct. Whether or not it actually is correct is up to a roll of the dice.
>
> Any knowledgeable AI user will tell you: AI *cannot* be safely used
> unless you already know what you are doing and can review the results
> before running it. AI proponents find AI useful to reduce the manual
> labor involved in writing the script -- ***not*** for figuring out the
> correct thing to do. This is an iterative process of rejecting answers
> that look wrong and tweaking the prompt, getting the AI to try again and
> again until it gets the right answer.
>
> DO NOT USE AI AS A TEACHER. YOU WILL BREAK YOUR SYSTEM IRREPARABLY.
>
> You will delete all your data by accident.
>
> You will be the latest example of a foolish fool posting on Reddit in
> tears, asking if there is any way to recover years worth of labor that
> wasn't backed up properly.
>
> You will learn nothing, or you will learn how to write incorrect and
> buggy scripts because you aren't able to evaluate which answers are
> hallucinations.
>
>
> --
> Eli Schwartz
<hehe>
I sense a certain overriding fear about the future...
</hehe>
As I said, lots of videos on YouTube supplemented by AI and tested
on copies of data or test machines. Additionally I did give some
guidance on how Dale might, if he was to use these tools, double
check the results before committing a grievous error on his
hardware. He could ask other AI's that did not produce the code in
question what it does. He could post it here for comments and
improvements.
Running scared is, to me, silly. Dale wants to learn. That's great
and these tools are one way to approach the problem. I'm using
them successfully every day of the week. Once I learned how to
ask questions correctly I found they don't hallucinate much, at
least with Python coding. I have built my own AI's for recognizing
playing cards. No way would I have made this progress without
the help I got from these tools.
Purchasing a book is a great thing to do, as I said before, if it is
recommended by someone trusted. (Or if you have money to
burn and shelf space for collecting dust after the 2 hours you use
it.)
To each his own. I agree with you in general - don't trust AI with
anything important. If you prefer then 'trust but verify'.
next prev parent reply other threads:[~2025-06-05 16:18 UTC|newest]
Thread overview: 24+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-06-05 3:49 [gentoo-user] Books about making shell scripts and other nifty commands Dale
2025-06-05 15:32 ` Mark Knecht
2025-06-05 16:02 ` Eli Schwartz
2025-06-05 16:16 ` Mark Knecht [this message]
2025-06-05 17:19 ` Eli Schwartz
2025-06-06 14:14 ` [gentoo-user] " Nuno Silva
2025-06-05 16:59 ` [gentoo-user] " Dale
2025-06-05 17:44 ` eric
2025-06-05 19:50 ` Re[2]: " Stefan Schmiedl
2025-06-06 2:58 ` Dale
2025-06-05 17:03 ` Jay Faulkner
2025-06-05 17:56 ` yahoo
2025-06-05 18:48 ` Arve Barsnes
2025-06-05 19:05 ` Eli Schwartz
2025-06-08 0:09 ` [gentoo-user] " Nuno Silva
2025-06-05 18:33 ` [gentoo-user] " yahoo
2025-06-10 9:39 ` Dale
2025-06-10 10:33 ` Quico Jurado
2025-06-10 14:33 ` Mark Knecht
2025-06-10 15:01 ` Eli Schwartz
2025-06-10 16:24 ` yahoo
2025-06-10 18:12 ` Dale
2025-06-10 18:07 ` Dale
2025-06-10 18:18 ` Quico Jurado
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to='CAK2H+ed88JaNrrHgZt5Bpw=y7zaucsmktgmgbJwGTGF0DAkm=Q@mail.gmail.com' \
--to=markknecht@gmail.com \
--cc=gentoo-user@lists.gentoo.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox