Artificial Intelligence, OpenAi... A way for Libre/Open Source software documentation and support

Hi, thank you for LO and for your time.

I am not used to say that, but it seems to me that Microsoft (and the others) have a serious head of us.

I have attended a video session on CoPilot, the AI implementation in all MS’s product.
It’s fantastic. And how to use OpenAI to get some answers on “how to” with your spreadsheet (and others). To be seen: https://youtu.be/Rq5nlJZPJXk

I started to use OpenAI to get some solutions with LibreOffice Calc. Fantastic.
See this thread: [Updated] Somme des cellules A1-A10

In the Free / Open Source software, it is always difficult to have time for writing documentation or for support. And in all languages. Especially towards new users.

Let me share this. I can imagine we don’t have the resources for doing the same thing.
But about documentation and support. We can image, for instance here, a button below your question. “Do you want to try an answer on our IA, train by TDF”.

This is not a suggestion or a proposition (I am not dev). Just interested in this topic.
This post on Mastodon :slight_smile: JRelland :linux: :flower: : « I am not used to say that, but it seems to me tha
 » - Framapiaf

Have a nice day. José from France.

I don’t get what you’re suggesting. If your goal is to facilitate discovery of solutions for questions relative to LO usage, AI could effectively help users to find similar problems. But there is a flaw: AI presently is only data mining and tentative extrapolation from what was found. This requires pre-existing data, i.e. at least documentation.

I’d say that probably as much as 75% of AskLO questions are the results of not reading the existing documentation and refusing the fact that LO usage departs from inherited routine. The remaining 25% are really interesting because they ask for “creative innovation”.

I agree that AI could easily solve the 75% which betray some laziness from users. I doubt AI could propose breaking suggestions in the latter case. From what I hear (I have not tried ChatGPT and consequently have no opinion), many AI engines only reformulate differently the question and the solution probably emerges from some new understanding of the problem by the user. This is anyway a plus but I don’t see the true added-value.

1 Like

You know, JosĂ©, I partly share your excitement about AI - it’s really impressive. Like the David Copperfield show, for example (you see that you are being deceived, but you agree to endure it further because it’s fun).

I’m talking about something else now. You publish a link to the results of your dialogue with the robot over and over again, but you don’t take into account that the session with the code 3b22ea98-f125-4216-9f2a-cf55fe2cfea1 is available only to you. All other users, when they try to use your link, end up on the Get started - Log in - Sign up start page. If you want to share exactly the result of the dialogue, then just publish the text, do not force us to register there.

2 Likes

Oups, sorry @JohnSUN about the link. My mistake.

I have updated the link: Somme des cellules A1-A10

Have a nice day. José

Yes, this is an amazing piece of intelligence. Let it be artificial, but at least some
 A living person would write “Press the Alt-Equals key combination so that a formula with the SUM (SOMME) function appears in the cell and select with the mouse the range of cells that should be summed - in your case A1:A10”

1 Like

Thank you for your joke @JohnSUN .

Did you see the following question in the thread? :slight_smile: About statistics?
Do you know a key combination which allows providing a solution about statistics :slight_smile:

And now, I am working on a request of Python program in Calc :)
 It does not work.
Let me tune my question :slight_smile:
Calc Sheet Tab Pop-up :slight_smile:

We may smile, joke, be serious for a while

I’m tempted to get very serious about all that, even more since I read a big deal of Max Tegmark’s “Life 3.0”. It’s not so much the stories he tells and the scenarios he presents. It’s about the current state of “global society” I found there concerning the ways of thinking that are presented as factually “common”.

When I (German highschool teacher M/Ph) had accompanied a group of students to Wisconsin for an exchange visit in 1992, and about a year later when a group, of US highschool students visited Bavaria in turn (and I accompanied them on occasions), I got the impression many of the US young had lost the ability to walk on feet through a city. They were used to cars, obviously.
Now I feel many people may have nearly lost the ability to use printed sources of “knwoledge”.
After occasional use of an AI based translator to check my English writing, I already often feel tempted (and sometimes I yield) to simply translate something from German to English instead of creating an English text myself - as I was used to for many years despite the doubtable results.

  1. Another 10 years from now: Who will still take the trouble to learn a foreign language?
  2. Who can tease my suspicions concerning the control and exploitation by global companies (their powerful forces) getting accesss to our complete global communication this way?
  3. Assuming a nearly complete breakdown of “modern” technical means (or a wilfully caused break of the functionality): How many of the currently > 8E9 humans would be able to make a living for another few years? In the end that may be a few even still capable of growing their grains and vegetables on some piece of soil. But long before it comes to the collapse of production most people will starve from missing traffic. Traffic concerning materials, but with a similar and faster impact of failing communication.
3 Likes

!!!it tells you nonsense about Sheet-tab-Popup!!!

I think there is one strong question for all AIs. How to fix a bugs in AIs if the output is bad?
Some AIs are really good in these days, for example e-translators. But there are always bugs in translations, but mostly not so terrible now. But there isn’t any possibility how to create really proper translator with AIs, because the indissociable part of AIs are the generators of “random” numbers. Without these pseudorandom generators, it is nearly unusable → because it always returns only one “strongest” variant.
I really don’t believe in illustrious future with statistical systems strongly based on “randoms”. “Randoms” always means there will be bugs. And how to repair bugs in large AIs? Put there a much larger set of some exceptions :-)?
Maybe easier could be to answer properly on Joiner paradox. Joke: Man who work with wood asks the new AI-IQ18000+: “I’m Joiner and I want to know the proper time when I must start to build own coffin. Because if I will start too much soon, there is danger the woodworms (anobiid) will eat it. And if I will start too much late, I will not finish it so it will be also work good for nothing.”


I don’t weigh to tip the statistics how many asks are from “lazy users” (because I don’t read all asks, mostly only ones with label ‘macro’), but I think @ajlittoz 75% seems well. But now, we talk about LibreOffice and how to solve the problems with LibreOffice, but what happen if there will be added the button for AI “solutions”? Will we must talk about unnecessary outputs from AI? Will it decrease the count of “lazy users”? Maybe in some cases. But the “laziness” have surely subsets that aren’t only “lazy” → “I haven’t so much time to read large (but really high-quality!) LibreOffice documentation”, “I’m frustrated from the vastness of documentation”, “I’m frustrated that I thought the solution will be easy and it isn’t so easy.”, “I spent a lot of time with documentation, but I still not understand fully to problems.”
So without AI we talk nearly about some live parts of our lifes, but with AIs it is only some dead mixture of some hard-primitivismed data.
For example I don’t loose my time with the asks that is something from MS Office, typically “I have XLSX file and I need” - I don’t know MS Office and I don’t want to learn how to work with paid/lease software. And I will not spent a time with the asks like: “Please help me, AI gave me this.”
My tip is the enthusiasm from AI will decreased (at some people), because there will start to discover in quantities the ugly bugs caused the “random” parts in algorithms. Without any possibility to fix it.


And I’m not able to imagine how the AIs will solve the mix of languages. I’m Czech and few weaks ago, some friends from Poland was here. We talked with mixture of Czech, Polish and English. Of course without any notices “this is Czech word, this is Polish, that is English”. We nearly normally speak the mixture of three languages and we proportionately understand to selfs :-), and it was amusing. And will be the mixture of languages easily distinguished for AIs? Easier for example than proper answer to Joiner paradox :-)? And what to talk some AI: “Do you know Joiner paradox? So write me what is Anobiid paradox! :-)”

Thank you for sharing your feeling about this topic.

Please note, I am using the free 3.5 version. I wonder what will be the answer with the 4.0 version.
And 5.0 version is already on the road.

See you soon. José

Version is not so important as solving the basic problem of inventing stuff.
.
We see in AI-generated code plausible looking code with functions wich are not available in the used API. This is also not acceptable for technical descriptions. So a human has to correct documentation. (Even now we see Mac-users complaining there is no Tools-Option entry in the Menu, as Apples guidelines tell to rename this.)
.
So we should not underestimate the work to do until you trained an AI to be really useful. And someone has to pay for the computing time for generation of the model and for running it to answer your questions also. This is easier for M$ than for TDF, even if they wished to do so.

1 Like

While AI might be useful to some, it’s a pain to others - I’m an attorney and I don’t like the idea of AI being trained on my data as there’s then an issue with the confidentiality of my data. I.e. if I’m writing a confidential letter in Word, I’m no longer confident that the content of that letter is confidential as it’s being sent to a LLM. If LibreOffice introduce AI, my one ask that it’s made very clear that it’s on and very clear how to switch it off.

1 Like

Hi @Richard_Gillespie, thank you for your time and I agree with your suggestion,

it’s made very clear that it’s on and very clear how to switch it off.

At the same time, the AI introduction here was more in connection with support and documentation.
Something like a “How to” function.

And your comment stay pertinent.

Enjoy with LibreOffice,
José from France.

It might be more in line with the goal of LibreOffice to consider making using of large language machines that run on users local computer or user’s server instead. Of course, large language models are computationally resource intensive and it might not align with everyone’s desire to run an LLM on their own machine, but I think there should be an option to allow so and to develop an LLM that would match the need and be easy to use to regular users of LibreOffice. But then there’s also another problem of would it take too much time/people/money to develop such an LLM for users of Libreoffice and would it be better to be a separate project that connect to LIbreoffice instead of for example a component of Libreoffice itself.

1 Like

[Upadete]
@Edmond DantĂšs :slight_smile:
(Intelligence artificielle géneratrice de texte. / Café Ubuntu / Forum Ubuntu-fr.org)

drawn our attention on this.

LibreOffice Content Generator with ChatGPT

Python macro on Github:

See the tutos from the author, Rania Amina:

See you soon and nice day :slight_smile:

Nothing related to LO found there


1 Like

instead posting dead links and links to random crap 
 @jrd10 should try to use OI (Own intelligence)

1 Like

@karolus :wink:

1 Like

@fpy: I know, but it wasn’t worth the attention.

1 Like

Sorry, it is not a dead link
just a mistake in the url


UPDATED :slight_smile:

@karolus @Wanderer @fpy