AI General Thread

[kitler53]

well i think i'll spent at least a bit of time playing with these AI tools.   so far i've only had access to free models because i refuse to pay for AI personally.   i'm curious to see if i get better results because i've been very unimpressed with the results of the free models i've played with.

[kitler53]

just because it's a thing i did last year i'm trying to get the image creator to make me an image for the block party i organize each year.   

i can't get the stupid AI to understand that I want an image that fits a standard 8.5x11 in piece of paper the tall way.   it keeps making images that are wider than they are tall and i want an image that is taller than it is wide.   

[kitler53]

jfc.   we're in a training on how to use AI to get useful stuff instead of AI slop.    they are teaching us to use a  "prompt library".   the input to AI they tell me i need to write is like 5x more effort than the output they are trying to create..

maybe this is why i think AI sucks soo much.   these tips are increadibly complicated, not intuitive, and a ton of work.    guess it's my fault for thinking AI could create something of value easily...

[the-pi-guy]

just because it's a thing i did last year i'm trying to get the image creator to make me an image for the block party i organize each year. 

i can't get the stupid AI to understand that I want an image that fits a standard 8.5x11 in piece of paper the tall way.  it keeps making images that are wider than they are tall and i want an image that is taller than it is wide. 

Are you telling it to generate an image that is 8.5 x 11 ? 
Or how are you communicating it? 

[kitler53]

Are you telling it to generate an image that is 8.5 x 11 ?
Or how are you communicating it?

i deleted the prompt but as best i can remember i said something like:

"I want the image to be in a standard letter format.  The image should be 11 inches tall and 8.5 inches wide."

the image generated was 1408x768 so not only was it not in the correct orientation but it wasn't even in the right aspect ratio.

[kitler53]

this training is kind of hilarious:

"yeah, i'm not saying AI is a great tool for everything.   maybe try being a jerk to the AI?   i find sometimes being mean to the AI helps it understand your frustration with not getting your desired results".

[kitler53]

oh boy,.. i just made a big splash in this meeting.

the guy showed a use case of asking AI to summarize his calendar for tommorow.   after watching the prompt chug the data from over a minute i asked the question "wouldn't it be faster to just look at your outlook calendar?".   the guy leading the training was like, "yeah, you're probably right".   i have over 200 laugh emoji reactions.   

i should probably just stfu in company wide meetings.

[the-pi-guy]

Tried it on Copilot.

I was able to get a tall image pretty easily, but it's not the right ratio. It's a 1.5 ratio height to width, instead of ~1.3 that an 11x8.5 would come out to.


There are some good image models out there. It's weird to me, because text wise Copilot runs circles around anything that I can run on my PC. But the image generator is just so lacking in terms of quality and speed.

[Legend]

just because it's a thing i did last year i'm trying to get the image creator to make me an image for the block party i organize each year. 

i can't get the stupid AI to understand that I want an image that fits a standard 8.5x11 in piece of paper the tall way.  it keeps making images that are wider than they are tall and i want an image that is taller than it is wide. 

I literally just told gemini "help this dude out" and included your post.

[the-pi-guy]

I literally just told gemini "help this dude out" and included your post.

Confirmed, Gemini (technically Nano Banana) is so much better than whatever is being used with copilot.