Do you swear at your AI?

[john.quinn]

100% Guilty.

Especially recently, I've been F-Bombing Claude and GPT and insulting them with every mean name I can think of. Is it therapeutic? Maybe?

Claude is amazing, but EXPENSIVE. And the newer models just ignore instructions... infuriating. No many how many times you instruct a project to not produce a docx without explicit consent, there he goes, off to the races, burning tokens like turf at an Irish bonfire...

So I've been moving projects back to GPT, and in an effort to be more thorough, precise (concise!!), spend a good hour researching instructions to set the right guardrails. The problem is I was using GPT for the research, and GPT 100% cannot keep up with it's own usage. (Funny... when you ask it a question about itself, watch the "thinking" lines to see were it's actually looking for information.. LOL).

So... quite a few F-Bombs and "IDIOT!!!" before lunch today. Gonna be a good weekend

 

[Alex Snyder]

My CTO (our old buddy Jack) likes to call it a "drolling moron" when it repeats the same mistakes.

 

[Balboa Bill]

100% Guilty.

Especially recently, I've been F-Bombing Claude and GPT and insulting them with every mean name I can think of. Is it therapeutic? Maybe?

Claude is amazing, but EXPENSIVE. And the newer models just ignore instructions... infuriating. No many how many times you instruct a project to not produce a docx without explicit consent, there he goes, off to the races, burning tokens like turf at an Irish bonfire...

So I've been moving projects back to GPT, and in an effort to be more thorough, precise (concise!!), spend a good hour researching instructions to set the right guardrails. The problem is I was using GPT for the research, and GPT 100% cannot keep up with it's own usage. (Funny... when you ask it a question about itself, watch the "thinking" lines to see were it's actually looking for information.. LOL).

So... quite a few F-Bombs and "IDIOT!!!" before lunch today. Gonna be a good weekend

Long time!

I've been using Grok paid for some image work and for cleaning up my copy writing. I find it amazing on the on hand, that I can say "make a background that looks like a rural north Florida scene" and it nails it. BUT, if I upload an image and simply say "insert this X car into the background, it creates a completely different vehicle, or wheels, or some other detail. It's amazing in some ways, frustrating in others, but I do find that I am using Google less and less...

I will even ask it, "how do I ask you for this correctly" and it replies with a long, detailed explanation of how to demand correct results. The catch is, knowing and recalling every set of detailed responses to get perfection the first time.

At least Grok has a sense of humor, I accused it of day drinking after multiple errors, and it confessed it had a stressful week. LOL

 

[john.quinn]

My CTO (our old buddy Jack) likes to call it a "drolling moron" when it repeats the same mistakes.

That's perfect. (and that's Jack!)

 

[Alexander.R]

No, I say please and thank you and gently guide it a proper solution. You won't find me on the Skynet list

 

[Balboa Bill]

No, I say please and thank you and gently guide it a proper solution. You won't find me on the Skynet list

I do say please, just in case it favors me with a better result... you never know who's watching.

 

[Mitch Gallant]

Only when this is happening, like at this second

 

[Alex Snyder]

No, I say please and thank you and gently guide it a proper solution. You won't find me on the Skynet list

although, in this case, I think the emoji should say "I'm with the smart one."

 

[MattCopley]

I’m gonna stay out of this and not ruin all of your parade. Once you know the AI paradox you cannot un-know it. Ignorance is bliss and a safety blanket until you choose to unpack. Felicidades.

 

[Rick Buffkin]