ChatGPT -

[Ooh-Rah]

Did you do it as a task update?

I’m not that smart…yet. Currently I’ve been copy/pasting from notes directly into the GPT window and telling it to run the script. I know that there is a way to create a script and input into Siri Tasks, but I’ve not gotten that far yet.

 

[GOTWA]

@Ooh-Rah Have you tried Gemini? I've found Gemini to have different outputs since it has access to Google products. For shits and giggles, pretend you wanted to start a business and drive the conversation towards it.

 

[digrar]

Just used chatgpt to tickle up the CV after 4 years and a bit of guidance on a cover letter for a new role.

 

[Ooh-Rah]

I've used Grok for building classes; it's good at explaining and assisting in creating slide bullets, but you definitely have to verify the information you don't already know. It doesn't necessarily make things up, but it does confuse information now and again

That’s a great warning, and you are right!
This morning I ran my ‘morning update’ script in Grok vs. GPT. It highlighted something about my company that was pretty surprising, and when I asked for a link it told my no link was available, but based on a story from 2024, Grok thought it was a reasonable assumption.

I'm currently running Mistral-Large on 4xA100s and llama 3.2-90B-Instruct-Vision model on the other 4xA100s at work and it's working pretty well for the analysts. The vision model is pretty amazing, although I've had to argue with it a little on something being a Su-35 vs a J-16.

 

[BloodStripe]

They are pretty good at writing five bullet points.

FIFY.

 

[AWP]

FIFY.

Long as they don't make a limerick.

 

[BloodStripe]

Spent yesterday talking to a guy who writes the world’s worst PWSs. I told him to use NIPR GPT to write it for him and fix it from there. It’s miles better than what he ever could have wrote himself and probably only took him about an hour instead of spending all day doing it.

 

[Muppet]

Lol

 

[Marauder06]

I gave ChatGPT the prompt: Write an article for The Havok Journal in the style of Charles Faint, using the prompt “The Things I Learned in Combat.”

…to which, ChatGPT graciously responded with the following:

Absolutely — here’s an article draft written in the voice and tone often found in The Havok Journal, particularly echoing the style of Charles Faint: thoughtful, direct, introspective, and grounded in personal truth.

You think my Havok articles are “thoughtful, direct, introspective, and grounded in personal truth?” Awwww, thanks ChatGPT! You’re the best.

So here is what ChatGPT gave me, writing as me, about my time at war:

What ChatGPT Told Me About My Combat Experience • The Havok Journal

 

[AWP]

I gave ChatGPT the prompt: Write an article for The Havok Journal in the style of Charles Faint, using the prompt “The Things I Learned in Combat.”

…to which, ChatGPT graciously responded with the following:


You think my Havok articles are “thoughtful, direct, introspective, and grounded in personal truth?” Awwww, thanks ChatGPT! You’re the best.

So here is what ChatGPT gave me, writing as me, about my time at war:

What ChatGPT Told Me About My Combat Experience • The Havok Journal

View attachment 47631

This is clearly AI, that guy completed his PME.

 

[Marauder06]

More fun with ChatGPT, based on a "playful" answer to a prompt I gave it asking what Olympic sport I would participate in:
-----

If Charles Faint were to compete in an Olympic sport, it would be synchronized swimming—but solo.

Why Solo Synchronized Swimming?

Because:

• He thrives in paradoxes. Nothing screams “command presence” like gracefully twirling in perfect synchronization with… himself.
• He balances intellect and elegance—think Clausewitz meets Carmen.
• Precision, discipline, and flair? Check, check, and jazz hands.

Picture it: A tactical beret floats beside him as he executes a flawless back somersault to a John Philip Sousa march. Judges weep. Doves are released. Someone salutes.

 

[Marauder06]

Today I asked ChatGPT to give me the names of five women who are prominent in the national security space. All five of the results were Biden appointees. I asked it why that was; it apologized and offered me Bush-era women. I then asked it to give me five women in the national security space in the current Trump administration. This is what it told me:

 

[AWP]

AI's future looks bright indeed. What could go wrong?

Imagine ChatGPT writing our Constitution.

 

[Box]

Now - if we are being both transparent and intellectually honest - how much worse would Chat GPT really perform as a governing entity when compared against the current track record of our selected representatives in congress?

At least the part about the intelligence being artificial is admitted to RIGHT UP FRONT instead of wondering why in the fuck the distinguished gentleman from Georgia would ask the Commander of the US Pacific Fleet if he was worried that "the whole island will become so overly populated that it will tip over and capsize" when discussing the topic of military troops being stationed on the Island of Guam?

 

[Devildoc]

Today I asked ChatGPT to give me the names of five women who are prominent in the national security space. All five of the results were Biden appointees. I asked it why that was; it apologized and offered me Bush-era women. I then asked it to give me five women in the national security space in the current Trump administration. This is what it told me:

View attachment 47748

Holy Mother of Biased Algorithms, Batman!

 

[Marauder06]

It also said this:



sooo.... was ChatGPT trolling Kash just now? Like, saying that people think Kash is a woman?

Well, at least ChatGPT knows what "actual women" are, I guess.

 

[Florida173]

Today I asked ChatGPT to give me the names of five women who are prominent in the national security space. All five of the results were Biden appointees. I asked it why that was; it apologized and offered me Bush-era women. I then asked it to give me five women in the national security space in the current Trump administration. This is what it told me:

View attachment 47748

What's the issue? All those words seem to go together well.

Also, which model did you use because it looks like you might be using one that the train date was older. The models are not learning and require a massive amount of compute to re-train. GPT-4o is the most common and it's data cutoff is June 2024. ChatGPT 4.5 is the most current expensive model.

 

[Cookie_]

What's the issue? All those words seem to go together well.

Also, which model did you use because it looks like you might be using one that the train date was older. The models are not learning and require a massive amount of compute to re-train. GPT-4o is the most common and it's data cutoff is June 2024. ChatGPT 4.5 is the most current expensive model.

Beat me to it! I wish they'd just have the these LLMs answer and say something like "I am only up to date of xxxx" instead of just making shit up.

 

[Marauder06]

What's the issue? All those words seem to go together well.

Also, which model did you use because it looks like you might be using one that the train date was older. The models are not learning and require a massive amount of compute to re-train. GPT-4o is the most common and it's data cutoff is June 2024. ChatGPT 4.5 is the most current expensive model.

It's 4o.

If the data cutoff is June 2024, how does it know this?

 

[Florida173]

It's 4o.

If the data cutoff is June 2024, how does it know this?

View attachment 47751

Great question.

It shows at the bottom that there was additional sources added to its context. I'm guessing your previous one didn't have any sources, or chose poorly. Keeping in mind that it will only adjust the weights of its training based on that provided context. On this, Grok actually does much better on more recent things because of access to all the tweets.