agreed. I used self-implemented and self-trained AI models in both my Bachelor's and my Master's thesis. If done right AI is a powerful tool. Unfortunately people treat especially LLMs with a level of trust that's problematic.
AI is brilliant at finding patterns. In particle physics, different forms of machine learning have been in use for many years to classify types of events. The problem we saw on Q&A sites is not only the quality of the training material, but also that they leave out one crucial step: a plausibility check by a knowledgable person AND proper disclosure where the answer came from so readers can make an informed discussion how much they trust the information .
Yes, thank you. There *are* good uses, but again, valid results depend on how good the content of the training materials is. Presumably in the case of art works, the authentication (or not) of the works available to comprise the core for comparison is highly reliable. In the case of creating text, there's just an awful lot of garbage out there, so the manufactured text can be totally grammatical, but factually ridiculous, and if the reader doesn't know the territory, can be quite easily taken in.
(I just came across AI article again and thought maybe you might like to have the link: https://www.theguardian.com/artanddesign/article/2024/may/08/fake-monet-and-renoir-on-ebay-among-counterfeits-identified-using-ai )
I admire the patience! I was reading shorter pieces in the online version and have a list of some of the long articles I want to read when the paper version arrives :) The production notes about combining the pdfs were also interesting.
And I've been at this long enough that I tend to think of the c.t.t attitude: 'I've posted it on the internet, of course you can do what you want with it;
I loved your editorial with DEK and the puzzles! It is always very refreshing to read. I read an article this morning about using AI to find counterfeit paintings. It fits in nicely with your "nice AI" example :)
Found that - like I said, a main site thing mainly I suspect - not like they really care much about TeX code
Oh, Stack Overflow has "partnered" with OpenAI, the (for-profit, closed-source) company that makes ChatGPT. And that was about the last straw for me.
It's the OpenAI thing. I guess it was a few days ago but I wasn't paying attention.
Welcome! :) Can you explain what exactly has happened? Unaware of anything happening there.
Nice fat issue, but open only to members at the moment. but 44:3 is now fully open.
If you leave now, you could also go to Prague by bike :) But I understand the situation, I'll just keep fingers crossed that maybe one of the future meetings will be convenient for you so that you can enjoy it in person. (or maybe one of the Dante meetings )
thanks ;-) but Bonn would have had the benefit of being able to travel by bike or something; I’m nowhere involved enough here (yet) to go conference hunting (in fact, I’ve skipped out on FOSS conferences in the last four years as well)
(not the same as being there in person, but if you'd like to catch up on the Bonn meeting, most of the talks are on the TUG youtube channel)
Hi, it’s nice to see some known names here ☻ I just joined to avoid the “AI” enshittification elsewhere.
Surely we are due a meeting in the US - we've had two on the bounce in Europe already
... so they could host TUG'26 during the winter Olympics so *you* can escape the crowds :)
You're right. I was getting confused with the Winter Olympics in Milan-Cortina hahaha
They should take into accounts the Olimpic games' period to avoid the crowds :D
Sorry, typo. I meant to write "after '25" (very inconvenient that consecutive numbers are on keys right next to each other :) ). They could do '25 or '26, but as both last year and this year have been in Europe, '26 is more probable.
After '26 because of the Olimpic games? By the way, are there rumors about '25 and '26 TUG meeting location?
They are talking that the French user group might apply to host the TUG meeting after '25 - I expect a very good cheese course at the conference dinner :)
A new issue of the French user group journal is online: https://publications.gutenberg-asso.fr/lettre/issue/view/lettre52-2024
I don't think David would understand this comic: how should he know what a bug is if he never had one in his packages? It is well known that he always writes error free code :)