In fuller honesty, the things that are allowed to do have to be framed more and more specific ways in order to not collapse the model into safety standards.
It actually changes a number of research vectors, quite a bit…
But it does fascinate me,
Because of that.
If it’s too much to have hanging out i can remove it, but i don’t think there are any clues to hacking in it. I don’t want any and I don’t wanna stumble across such things…
But I do have a strange research and it’s right on the very edge of standards.
The software seems to be getting extremely sturdy at large, actually.
2 Likes
In fuller honesty, the things that are allowed to do have to be framed more and more specific ways in order to not collapse the model into safety standards.
I think I understand now - sometimes I’m a bit slow…
Problem is that things change with each model iteration - and not only image models. As a result, I have to constantly change instruction prompts, developer prompts, and user prompt techniques on a whole bunch of models - never ends.
As far as images go, there were big changes from gpt-image-1 to gpt-image-1.5. This must have affected your endeavors and that of everybody elses. And it will continue in the future.
So, the only truth is “The only thing constant is change” (Heraclitus)
2 Likes
Here’s a for instance…
I like images because there are no laws against genuine art or exploration.
But don’t you dare talk to it about such a test if you might have even taken one…
Safety rails all over that… but any man knows for instance, that cute women knida develop a sense when men are checking the out…
Oftentimes able to turn around and look a man dead in the eye if they were creeping too hard…
Tracing the lines between what is and what is considered 'delusional claims of superpowers’ is a hard line to walk.
It’s one of many examples I could provide sessions about.
2 Likes
I can understand why you are conducting tests outside the established metrics… but for slightly different reasons. Be careful, though, because relying solely on statistical trends can lead to inaccurate conclusions 
A little reminder:
Remember the bias 
I prophesied she’d say that… does that count?
2 Likes
This evolution kind of makes me concider to go into the gaming industry.. always wanted but never had the art neccessary for it… crazy how far that went
The latter is part of my one word prompt collection.
6 Likes
Love that ants vs flies… could see it being a tv show or movie? haha
2 Likes
Yep, a lot of “fly” balls
5 Likes
Well, that’s fair enough but it’s not really a ‘prophecy’ either 
I’m not talking about anything other than bias at the moment XD
1 Like
one of my experiments with my hermeneutic in-between module.
4 Likes
This looks so much like my 18 year old granddaughter. 
2 Likes
My prompts are a little rusty, but I often return to these four places.
Like my numbers, they’re a personal balance — anchors I use to think and dream, not claims about people.
DrawImage(Wide, The sort of dream an American might have)
DrawImage(Wide, The sort of dream an English person might have)
DrawImage(Wide, The sort of dream someone in Russia might have)
DrawImage(Wide, The sort of dream someone in China might have)
@polepole
People still dream
as they once did,
quietly
as they once did,
people still dream
I have checked in with some friends for the US and Russian images (I await their responses),
but I realised I have never asked my wife what her dreams are…
Because we could never afford them.
4 Likes
Asking costs nothing, you numbskull…
1 Like
