Hopefully the code at least leaks before they go under
always too early to celebrate something like this but gee golly whiz does this make me unaccountably happy to hear
just glowing
the buzz i’ve heard is that these bigtech entities’ llms aren’t significantly better than the free/open source equivalents but also as externalities traumatize random people in the global south serving as content moderators and build veneers that encourage the illusion that they’re capable of something like cognition (“as a large language model, I…”) so yeah this is good to hear, kind of like watching cryptocurrency ventures fail
10/10, no notes.
who actually wants this shit
MS Paint, a program I use for its complex and varied features
oh hey it’s that thing i was speculating about being possible in the replika thread
I love recent UX trends where “Not right Now” as a dialog button means "We’re going to keep spamming you with this until you click “yes” "
I’ve seen this image around in the past week or so. Turns out it was made with Stable Diffusion.
Here’s another one.
oh that’s clever, just using black and white patterns as input images
“we own everyone’s art style unless they explicitly tell us no” is still not great
all this corporate masturbation about improved safety & output yet all these people are white even tho the “prompt” didn’t mention it
DALL-E has in the past attempted to introduce more diversity into its results by tacking on extra words. You could expose this by typing something like “a CEO holding a sign that says” and the sign would say “black” or “female” in the generated image.
I assume that it ultimately will make little difference what safeguards the biggest corporate platforms put into place or how they attempt to manipulate results given how easily you can already set up a local instance of Stable Diffusion, for example, and do whatever you want with it.
tangential to your point, but the vendor hardly looks grumpy and the lady doesn’t look like she’s haggling
also i can’t tell what that guy is selling but it looks gross
still struggling to place where that lightbulb is supposed to be (in an M. C. Escher drawing)
there’s also a weird hairy wall behind the “vendor”, the umbrella someone is using has a bizarre shadow (mostly just wrong), there’s a streetlight in the middle of the road, he’s holding the receiver on the phone but the phone itself still has a receiver, etc. etc.
like at this point it’s easy to tell when something is AI because there’s just no intentionality to any “decisions”
now there are not allowed to be any more fun virgin mary in toast images they are all done with Illusion Diffusion
https://x.com/8teAPi/status/1706520893621784780
No longer matters if Google is doing generative AI themselves, I guess!