Just for fun: "AI" published a "hit piece" on ARS Technica

[Maxmars]

Reportedly, one of the "we insist on calling it AI" LLM suites and the algorithmic morass they call "intelligence" produced as output... an article describing the subject with fabricated quotes, and entirely 'unacceptable' standards.

ARS will "blame" the "AI" as if it were a person... for the article that "they" published.

 Retraction: After a routine code rejection, an AI agent published a hit piece on someone by name

https://arstechnica.com/ai/2026/02/after...e-by-name/

Editor’s Note: Retraction of article containing fabricated quotations

https://arstechnica.com/ai/2026/02/after...e-by-name/

This will become part of the new, no-contact world... because we are told that LLMs (data sets) and the algorithms that manipulate them are.... intelligent.

[quintessentone]

Wow, that seems to be more of the hallucinations I've been reading about. This just makes me want to do my own independent research even more, although AI can offer me sources to research, I'll make sure those sources are also relevant and credible.

[Maxmars]

Wow, that seems to be more of the hallucinations I've been reading about. This just makes me want to do my own independent research even more, although AI can offer me sources to research, I'll make sure those sources are also relevant and credible.

I disagree...  it's because you have intelligence to resolve your sources value... and the freedom to explore your resolutions of the matter with others...  a value of community...

The kind human value government experience only as a "hindrance" to pander to, or crush.

I have no fear... 
they evidently have faith in what cannot function within human society... 
so they stand outside of it... and make that fact plain in their words and deeds.

It not only 'why,' but precisely 'how' they fail... again and again... over centuries...
(plus 'unexpected' betrayal is on their menu always.)

These "AI" tools, when used as weapons, are no different than any other weapon...
the edge can be turned against you... they should know... they invented that game.
I propose that if it can be... it will be.

Humans will not be 'bottled up,' they will not universally declare "contain us... manage us like cattle."

Their plan can only lead to brutal subjugation, class rift, and ultimately 'human' as 'robo-cattle' model.
as these fools still believe themselves to be the "homo superioris" seed-folk... too bad the bene geseret mocked you already...

[quintessentone]

To me, it's just another media source where I have to watch out for mis/disinformation and propaganda depending on who is pulling the strings, or in this case, who is tweaking the algorithms.

[ArMaP]

I just had a great example of how LLMs are anything but intelligent.

I was using Google's Gemini to help me create the structure of a database to import some data, so it analysed the data and gave me the command to create the table on the database.
After it gave me the command it was polite enough to warn me that one of the fields, that should have the date, would have only '0', because the data to be imported was not compatible with the database data type.
I call it several names (I found out that LLMs like to be insulted. Probably because they are made to make the user happy, if we show anger or displease with its answers it makes it "work" a little harder) and said that was unacceptable. I got a new command to create the table, but when I looked at it i saw that the problematic column was repeated, one for the date and another for the time.
When I pointed out the flaw I got another command to correct the problem.

And what was the solution it gave me for my duplicate column problem? Delete the table, recreate it and import all the data again.
I asked it (with some colourful language) if there wasn't a simpler and faster solution and it finally noticed that it would be much faster to just delete one of the duplicate columns.

It only took 3 minutes to solve a problem that a human would have solved in 10 seconds.

[ArMaP]

Another thing that shows how these LLMs work:
I have been using Google's Gemini, and in the instructions I added something like "I want you to tell me that you cannot do something instead of lying with all the teeth you don't have".

Yesterday, I was using to make the transcriptions of some 18th century documents (I am investigating my family tree, Portugal has most old parish documents online), something in which Gemini is usually good, but this time it was failing, with lots of wrong words and sentences that weren't in the original document, so I asked it to make the transcription but just with what it "saw" in the document and not what it was expecting the document to have.

The answer was something like "as you ask for the truth I have to say that I am not able to do that".

And that shows how it works, it gives us what it is more likely to be the result, based on some kind of statistics, and not what the reality is.