Perhaps, depending of context, it doesn’t always understand OS and in some places writing operating system would help.
If you were trying to obtain the same kind of picture for a M$ OS, you would probably give the version. What if you say that want to see Haiku beta4?
1 Like
Flux outputs are less artistic, but it never writes Haiku ΟS. It also has a strong bias “computer = mac”.
Anyway, that’s all I wanted to know. For doing actual art there’s a better tool, it’s called 3D modelling.
1 Like
Because you asked it for a picture of “the OS”
I stead try writing operating system and spell out what it should write.
“The screen should show the text Haiku” etc
1 Like
I think that it doesn’t see Haiku as a reference it has to use. So it creates something based uniquely on your words.
I would rather phrase it like that. “Create a gouache painting with nice colours representing a computer running Haiku operating system, with a flat monitor, a keyboard and, a mouse. On the screen, the desktop wallpaper must be the Haiku operating system logo.” and see if that helps.
Ultimately, I think most current so-called AI models just simply do not know or even understand what Haiku the operating system is. Then again, the most popular type of ML these days is LLM (aka repeating words). There is no form of understanding going on at all.
Just for fun, I asked ChatGPT to write me a tiny CLI app with just one functionality.
Didn’t compile at all.
1 Like
Pattern recognition is arguably a form of understanding, even if a primitive one.
Try asking it for a table with no computer on it.
You’ll see the problem.
1 Like
That looks like Elementary OS.
2 Likes
Looks like it tried to make Haiku the monitor brand name but somehow didn’t manage either.
The amazing thing is that more we try, smaller and more compact are keyboards. The screen is still flat but for the first time we have the computer unit on the desktop. I can’t get what is the device between the unit and the screen… AI seem to all have problems with wiring things together.
