Re: Is [email protected] still active?

ThreeBlindQuarks via Tutor <[email protected]> Mon, 07 Jul 2025 18:54:26 +0000
Newsgroups gmane.comp.python.tutor
Message-ID <IcVRW3qxvlcCAm9f5Sn41urNG_9vb7uGTDxKdUawlyzy5aanfb9W87X01ZBNkbbQNJocGjeiX1qslig0H7dkZ4d-tN1npLrCwsA8KsZvbt8=@proton.me>
That is an excellent point, in general. The current generation of what they call AI needs training data.

Imagine a brand new language with initially just some manuals with limited examples. Not much public discussion as to tricks and techniques and warnings and limitations and even lots of bugs in the code and documentation can lead to a horrible AI.

Even with existing languages that change, a question may result in answers that do not apply to your version or suggest modules that lost support or have made major changes.

I recently read about a change in another language called R that added various forms of objects that were far less robust or flexible than in python. Most older packages were written using S3 objects and some newer ones used S4 and they now created a sort of upward extension called S7. You an still use any variety but one goal was to gradually move many old things to S7 and that can be a major rewrite even if the external interfaces remain mostly the same.

I now see warnings that a much-used graphics package is being replaced by an S7 version. Many separate packages exist that build on or extend it and it has been noted some such as the bioconductor can fail as they extend qualities of graphs. Users are warned not to upgrade until they have tested against it with their own code.

Back to the point is what would an existing AI know? Would one already trained get new info in time? Can it ask the user to specify more detailed info such as what version numbers it is using under what OS and so on? Worse, some complex programs can look like they worked but produce wrong results. This is true even with human help. 

One related issue is that many users do not take care in programming such as checking for error situations and may even end up destroying their data. Imagine reading in data from a file and applying some transformations and overwriting it.  If the code from an AI is not a match, the result on breaking can be bad. Novices with minimal understanding may not consider much ... and not provide detailed guidance and GIGO applies.



Sent from Proton Mail Android


-------- Original Message --------
On 7/6/25 2:21 PM, Alan Gauld via Tutor <[email protected]> wrote:

>  On 06/07/2025 13:05, Mats Wichmann wrote:
>  
>  > from AI, why bother with a mailing lists?
>  
>  Well, somebody has to give the AI the answers it regurgitates! :-)
>  
>  --
>  Alan G
>  Author of the Learn to Program web site
>  http://www.alan-g.me.uk/
>  http://www.amazon.com/author/alan_gauld
>  Follow my photo-blog on Flickr at:
>  http://www.flickr.com/photos/alangauldphotos
>  
>  
>  
>  _______________________________________________
>  Tutor maillist  -  To unsubscribe send an email to [email protected]
>  To unsubscribe or change subscription options:
>  %(web_page_url)slistinfo/%(_internal_name)s
>  
_______________________________________________
Tutor maillist  -  To unsubscribe send an email to [email protected]
To unsubscribe or change subscription options:
%(web_page_url)slistinfo/%(_internal_name)s