[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [ELPA] New Package: minuet-ai.el
From: |
Philip Kaludercic |
Subject: |
Re: [ELPA] New Package: minuet-ai.el |
Date: |
Sun, 16 Mar 2025 00:17:12 +0000 |
Henrik Ahlgren <pablo@seestieto.com> writes:
> <tomas@tuxteam.de> writes:
>
>>> (Btw, AFAIU this package is specially tuned to use code-generating
>>> models, and those are a far cry from the ones you call "bullshit
>>> generators", which produce human-readable text with "expert
>>> opinions".)
>>
>> They are exactly the same LLM architectures, used on code. And they
>> "hallucinate" (their term) or "bullshit" (our term) in very similar
>> ways. Their business model is producing code which "looks" correct,
>> just as their cousins do with lanuage. Arguably, this is even slightly
>> more dangerous.
>
> In my experience, most LLMs perform rather poorly at code generation,
> especially with less mainstream languages like Emacs Lisp. Obviously, I
> haven't tried every model, and I also recognize that prompting skills
> significantly impact the outcomes.
@Tangent{I would actually be interested to hear if minuet-ai.el (or
whatever the final name is) was written with AI and to what degree.
Reviewing the code, it was not so much obviously buggy, but there were a
number of unusual choices made that I haven't seen before, or at least
not as often. It might be worthwhile to collect the kind of "soft"
issues that LLMs make when generating Emacs Lisp code, to have some
checklist when people decide to use LLMs when writing, aiding to write
or learning to write Elisp -- or perhaps even as a useful hint for the
LLM itself?}
>
> For me, the most beneficial use case for current generation LLMs is as
> an advanced spell checker, particularly when writing in English, my
> third language. I believe my computing freedom is not significantly
> compromised by using a SaaSS to enhance the language of my writing.
>
> I ran this piece of text through an LLM: since this is public content, I
> have no privacy concerns, and once I finish, I don't mind if the LLM
> provider goes out of business and deletes all my data (I would actually
> prefer that). And the cost is a fraction of a cent.
- Re: [ELPA] New Package: minuet-ai.el, (continued)
- Re: [ELPA] New Package: minuet-ai.el, Eli Zaretskii, 2025/03/10
- Re: [ELPA] New Package: minuet-ai.el, Philip Kaludercic, 2025/03/11
- Re: [ELPA] New Package: minuet-ai.el, Eli Zaretskii, 2025/03/12
- Re: [ELPA] New Package: minuet-ai.el, Milan Glacier, 2025/03/12
- Re: [ELPA] New Package: minuet-ai.el, Richard Stallman, 2025/03/14
- Re: [ELPA] New Package: minuet-ai.el, Henrik Ahlgren, 2025/03/15
- Re: [ELPA] New Package: minuet-ai.el, Milan Glacier, 2025/03/15
- Re: [ELPA] New Package: minuet-ai.el, Eli Zaretskii, 2025/03/15
- Re: [ELPA] New Package: minuet-ai.el, tomas, 2025/03/15
- Re: [ELPA] New Package: minuet-ai.el, Henrik Ahlgren, 2025/03/15
- Re: [ELPA] New Package: minuet-ai.el,
Philip Kaludercic <=
- Re: [ELPA] New Package: minuet-ai.el, Milan Glacier, 2025/03/15
- Re: [ELPA] New Package: minuet-ai.el, Eli Zaretskii, 2025/03/16
- Re: [ELPA] New Package: minuet-ai.el, Jean Louis, 2025/03/17
- Re: LLM-generated code in minuet-ai.el (was: minuet-ai.el), Eli Zaretskii, 2025/03/22
- Re: LLM-generated code in minuet-ai.el (was: minuet-ai.el), Milan Glacier, 2025/03/24
- Re: LLM-generated code in minuet-ai.el (was: minuet-ai.el), Eli Zaretskii, 2025/03/24
- Re: [ELPA] New Package: minuet-ai.el, Jean Louis, 2025/03/17
- Re: [ELPA] New Package: minuet-ai.el, Eli Zaretskii, 2025/03/15
- Re: [ELPA] New Package: minuet-ai.el, Richard Stallman, 2025/03/27
- Re: [ELPA] New Package: minuet-ai.el, Richard Stallman, 2025/03/13