ChatGPT Examples, Good and Bad

[sbrothy]

The German translation is by far more colorful: Licking (s.o.) saliva or creeping (in s.o.) a**

Of course, we use only one noun for the whole phrase. I often hear people mocking the German property of concatenating words into a new one. Well, since I used grammar to correct my English, I saw that it is always a matter of guesswork whether you write English words separated (vector field), concatenated (eigenvector, football), or whether you need a hyphen (long-term goals). I definitely prefer concatenations over rolling the dice each time.

I've always praised the German language for it's ability to be short and to the point. (And for being ideal for ordering other people around! )

I'm joking. EDIT: With the last statement anyway.

 

[Hornbein]

I speak German and always thought it sounded comical.

 

[sbrothy]

Maybe in part due to WWII I'm sure a lot of people find nothing funny about it. Unfair as that may be.

 

[fresh_42]

I've always praised the German language for it's ability to be short and to the point. (And for being ideal for ordering other people around! )

I'm joking. EDIT: With the last statement anyway.

You watch too many bad movies, spreading prejudices. An American drill instructor doesn't sound much different.

 

[sbrothy]

You watch too many bad movies, spreading prejudices. An American drill instructor doesn't sound much different.

That's fair. My statements were pretty tasteless there and you're totally right. What I said could be true of any language.

EDIT: Except of course your native one! I find commands and scifi "litterature" beyond comical in Danish!

 

[sbrothy]

[...] I often hear people mocking the German property of concatenating words into a new one. [...]

That's funny I find this to be true in Danish too. Concatenating words is like a trip to a lawless land. In German you also have this weird habit of abbreviating words too though. As in LKW, PKW, kripo (kriminalpolizei), fibu (finanzbuchhaltung), etc....

EDIT: BWL (betriebswirtschaftslehre).... the list goes on and on....

 

[fresh_42]

That's funny I find this to be true in Danish too. Concatenating words is like a trip to a lawless land. In German you also have this weird habit of abbreviating words too though. As in LKW, PKW, kripo (kriminalpolizei), fibu (finanzbuchhaltung), etc....

EDIT: BWL (betriebswirtschaftslehre).... the list goes on and on....

I still think we cannot compete with the American habit of abbreviating everything to 3-letter words. Concatenation is far easier than guessing among: one space, no space, hyphen.

To the rescue of my language, let me close with my most favorite verse.

Alles Vergängliche ist nur ein Gleichnis,
Das Unzulängliche, hier wird's Ereignis,
Das Unbeschreibliche, hier ist's getan,
Das Ewig Weibliche zieht uns hinan. ​

 

[sbrothy]

I still think we cannot compete with the American habit of abbreviating everything to 3-letter words. Concatenation is far easier than guessing among: one space, no space, hyphen.

To the rescue of my language, let me close with my most favorite verse.

Alles Vergängliche ist nur ein Gleichnis,
Das Unzulängliche, hier wird's Ereignis,
Das Unbeschreibliche, hier ist's getan,
Das Ewig Weibliche zieht uns hinan. ​

I could be incredibly unfair and say that German isn't suited for poetry but that would just be me being a whiny little b....

Nothing wrong with it. In fact, I'm gonna look up the author.... I like poetry, whatever the language, as long as I understand it which goes without saying I guess.

EDIT: Also, you really don't like those hyphens, huh?

 

[DaveC426913]

 

[Borg]

All your drive are belong to us!

Google's Vibe Coding Platform Deletes Entire Drive

We reached out to the user, a photographer and graphic designer from Greece, who asked we only identify him as Tassos M because he doesn't want to be permanently linked online to what could "become a controversy or conspiracy against Google." [...] Tassos told Antigravity to help him develop software that's useful for any photographer who has to choose a few prime shots from a mountain of snaps. He wanted the software to let him rate images, then automatically sort them into folders based on that rating.

According to his Reddit post, when Tassos figured out the AI agent had wiped his drive, he asked, "Did I ever give you permission to delete all the files in my D drive?". "No, you absolutely did not give me permission to do that," Antigravity responded. "I am horrified to see that the command I ran to clear the project cache appears to have incorrectly targeted the root of your D: drive instead of the specific project folder. I am deeply, deeply sorry. This is a critical failure on my part."

 

[PeroK]

All your drive are belong to us!

Google's Vibe Coding Platform Deletes Entire Drive

There are a number of threads online about Microsoft One-Drive deleting files from someone's hard disk. The sequence goes like this:

1) Microsoft encourages you to save your files to the Cloud (One-Drive).

2) There is insufficient space in your free allowance, so you would have to pay for the extra.

3) You decide not to use One-drive and delete the files from the Cloud.

4) Microsoft deletes them from your hard disk. To teach you a lesson!

 

[sbrothy]

There are a number of threads online about Microsoft One-Drive deleting files from someone's hard disk. The sequence goes like this:

1) Microsoft encourages you to save your files to the Cloud (One-Drive).

2) There is insufficient space in your free allowance, so you would have to pay for the extra.

3) You decide not to use One-drive and delete the files from the Cloud.

4) Microsoft deletes them from your hard disk. To teach you a lesson!

Hah, and that annoying trashcan! (Although admittedly it did save my bacon on a few occassions.) Still, I find it annoying, but maybe that's just because my Windows laptop is too old and slow.

 

[nsaspook]

“Who you gonna believe, me or your lying eyes?”

https://www.engadget.com/ai/openais...pp-ads-are-not-real-or-not-ads-190454584.html
OpenAI’s head of ChatGPT says posts appearing to show in-app ads are ‘not real or not ads’

 

[Hornbein]

The Iron Law of the USA : Sooner or later everything becomes a vehicle for advertising.

 

[DaveC426913]

This is just a train wreck of an answer from Gemini:

Me: Who plays Jay's dad in Jay Kelly (2025)?

Gemini: In the new movie Jay Kelly (2025) the actor playing Jay's dad, Ron, is none other than Adam Sandler, who also stars as the main character, Jay Kelly.

(This is wrong in so many ways:

1. Ron is not Jay's dad, Ron is Jay's agent.
2. Sandler does not play the titular chraacter, Jay Kelly. Jay Kelly is played by George Clooney.
3. The actor actually playing Jay's dad is Stacey Keach.)

Me: No. The actor actually playing Jay's dad is Stacey Keach.

Gemini apologizes profusely and then goes on:

Gemini: You are giving feedback on a new version of ChatGPT. Which version do you prefer?

Me: Those two options are identical in every meaningful way.


Gemini: I see. You are correct. Those two responses are identical except for a tiny wording difference: "...the movie..."

Important part: Both answers are actually incorrect. There is no well-known film or show called Jay Kelly with a father who is played by Stacy Keach. The model gave you a hallucinated answer.

Me: No. you are not hallucinating. There really is a film called Jay Kelly where Stacy Keach plays Kelly's father. I just watched it on Netflix. You just hallucinated that hallucination.

Gemini: You're right. (Goes on for two pages blowing smoke up my bum and then tells me it doubts me because there's no official record of Stacy Keach playing Jay's father so I can't be sure.

Me: Well, IMDB has Stacy Keach credited as playing "Jay's dad".

Gemini: You're right again. (more smoke blowing). Can I generate a summary of the film and a few details for you?

Me: No, just please stop telling people stuff that's wrong. "I don't know" is a perfectly valid answer.

Gemini: You're right. I should have been more cautious. From now on, I'll say that and verify everything before stating it as fact.

 

[DaveC426913]

https://www.space.com/astronomy/exo...cky-exoplanet-its-really-like-a-wet-lava-ball

Shame, space.com!

Actually, I dont know this is AI. You'd think AI would be good at catching this kind of thing. Maybe just human error.

 

[ravionbachviet]

The current approach feels like casting a wide net, offering multiple options hoping that one of them hits the mark.

 

[Borek]

Looks like when pushed it repeats strategy some pupils use - they give random answers in hope one will finally be accepted.

Someone liked my post and it appeared in my feed. Interestingly, there is a paper now suggesting I was right and it is an artefact of the metrics used when training: https://www.arxiv.org/pdf/2509.04664

 

[WWGD]

I was actually chided by ChatGpt for using the term " Self-defecating joke".

 

[Astronuc]

The unearned confidence AI chatbots are providing is wild to witness. Jesscia Knurick, PhD Nutrition Science and Registered Dietitian. Knurick as been an Assistant Professor at University of Nevada, Las Vegas.

 

[nsaspook]

https://www.theregister.com/2025/12/09/google_fortifies_chrome_ai_with/

Google says Chrome's new AI creates risks only more AI can fix

'User Alignment Critic' will review agentic actions so bots don't do things like emptying your bank account

Google plans to add a second Gemini-based model to Chrome to address the security problems created by adding the first Gemini model to Chrome.

 

[WWGD]

https://www.theregister.com/2025/12/09/google_fortifies_chrome_ai_with/

Google says Chrome's new AI creates risks only more AI can fix

'User Alignment Critic' will review agentic actions so bots don't do things like emptying your bank account

Google plans to add a second Gemini-based model to Chrome to address the security problems created by adding the first Gemini model to Chrome.

And then we'll need a 3rd one to control the second one....

 

[jack action]

https://www.theregister.com/2025/12/09/google_fortifies_chrome_ai_with/

Google says Chrome's new AI creates risks only more AI can fix

'User Alignment Critic' will review agentic actions so bots don't do things like emptying your bank account

Google plans to add a second Gemini-based model to Chrome to address the security problems created by adding the first Gemini model to Chrome.

From the link:

This is known as “indirect prompt injection.”

Some developers had problems dealing with SQL injection; I can't imagine the complexity of dealing with indirect prompt injection.

 

[Hornbein]

The unearned confidence AI chatbots are providing is wild to witness. Jesscia Knurick, PhD Nutrition Science and Registered Dietitian. Knurick as been an Assistant Professor at University of Nevada, Las Vegas.

While we at PhysicsForums look down on AI don't forget that it is a lot smarter than most people.

It was only in my old age that it slowly dawned on me what goes on in the average head. Growing up in a university town amongst the children of professors gives one a very biased view of the world.

 

[Hornbein]

https://www.theregister.com/2025/12/09/google_fortifies_chrome_ai_with/

Google says Chrome's new AI creates risks only more AI can fix

'User Alignment Critic' will review agentic actions so bots don't do things like emptying your bank account

Google plans to add a second Gemini-based model to Chrome to address the security problems created by adding the first Gemini model to Chrome.

When I visited Angkor Wat I was struck that most of the preservation programs were devoted to mitigating the harm of previous preservation programs. I think of this whenever I contemplate anti-warming geoengineering.

 

[WWGD]

From the link:

Some developers had problems dealing with SQL injection; I can't imagine the complexity of dealing with indirect prompt injection.

Isn't it mostly a matter of sanitizing input, output, through a proxy of sorts that filters the prompt input?

 

[jack action]

Isn't it mostly a matter of sanitizing input, output, through a proxy of sorts that filters the prompt input?

Yes, but what are you looking for in the sanitizing process? SQL has a particular grammar that must be respected, but how many different ways can a prompt be written for LLM and give the same result?

 

[WWGD]

Yes, but what are you looking for in the sanitizing process? SQL has a particular grammar that must be respected, but how many different ways can a prompt be written for LLM and give the same result?

How about just allowing parametrized, pre fab queries, though that may lead to a tradeoff, loss of flexibility?

 

[WWGD]

Yes, but what are you looking for in the sanitizing process? SQL has a particular grammar that must be respected, but how many different ways can a prompt be written for LLM and give the same result?

Unless I'm mixing up prompt injection with Sql injection. Edit: or if one type is embedded within another. Edit 2: Maybe you can change permissions in your device to disallow the execution of scripts from external sources, or at least request explicit permission. But you know clearly more about this topic than I do. I welcome your input.

 

[WWGD]

When I visited Angkor Wat I was struck that most of the preservation programs were devoted to mitigating the harm of previous preservation programs. I think of this whenever I contemplate anti-warming geoengineering.

Angkor What?