Oh, and what I tried to create with some of the CLIP-guided notebooks, but it never worked out: “A painting of a man-o-war chasing down a smuggler guns blazing”
Ah, unfortunate. But thanks for trying. Maybe without the guns blazing?
Dall-E also wasn’t brought up on a diet of chess problems, it seems. I probably should have specified to show a diagram. These boards and positions are positively wild.
The original DALL-E was capable of having almost the same image with slight variations in one generation, so I’d be interested to see something like “A photograph of a village in 1900 on the top, and the same photo colorized on the bottom”.
Book cover for “Using Credence Calibration for Everything”
If it’s possible to specify height and width, with the width being three times the height to be used for my LW sequence. Otherwise, I’m still interested in how this more abstract concept gets translated.
I can post a few more non-cherry-picked generations here, if there are some captions / prompts anyone would like to try.
edit: alright, I should try to get at least a little bit of work done today. might come back to generate some more later
Illustrated instructions for a playground game called “Less Wrong”—which will be featured in the TV series Squid Game 2.
An Ikea instruction manual showing how to build a superhuman artificial intelligence.
The “Game of Life” board game, updated for the year 2142.
A map showing Russia’s planned invasion of Ukraine, with key objectives circled in red.
A still image of the long-delayed Harry Potter sequel, “Harry Potter and the Reloaded Revolutions” coming out in 2028.
Tessellation by M. C. Escher
Interesting, thank you!
I’d be interested in whether it can count, so something like “Seven hedgehogs playing musical chairs”.
Cool idea. It seems that it isn’t great at this, as only a few of the generations had exactly 7, but they are mostly all around there (6 or 8):
Awesome, thanks a lot!
How about “White to mate in 3 moves”?
Oh, and what I tried to create with some of the CLIP-guided notebooks, but it never worked out: “A painting of a man-o-war chasing down a smuggler guns blazing”
Unfortunately, the 2nd prompt violates the content policy (violence)
Ah, unfortunate. But thanks for trying. Maybe without the guns blazing?
Dall-E also wasn’t brought up on a diet of chess problems, it seems. I probably should have specified to show a diagram. These boards and positions are positively wild.
The original DALL-E was capable of having almost the same image with slight variations in one generation, so I’d be interested to see something like “A photograph of a village in 1900 on the top, and the same photo colorized on the bottom”.
It did a fairly decent job, though it apparently prefers the colorized versions being on top.
Note that getting DALL-E to work well is a bit like GPT-3: you can usually get much better results by making a few iterations on the prompt.
Very cool, thanks!
Let’s have fun with recursion!
A checkerboard where each square is itself a checkerboard.
A cube with mirrors on both sides, the mirrors show multiple reflections of the cube.
A person wearing a shirt with an image of that person wearing that shirt.
My mom wants to see what this would look like: “An ant on a scooter on an interplanetary flight”. Let’s see how many concepts it can combine at once.
I could see your mother having a promising career in QA
Thank you!
“Prince Valiant fighting Batman” (I am kind of waiting for models to allow me the creation of comics)
“Einstein writing his famous equation on a blackboard”
Poor Einstein has dementia. And Dall-E doesn’t know the Prince Valiant comics. Still, this is so cool! Thanks again!
Plot of global infant mortality rate versus time.
I’d be curious what happens for the prompt:
If it’s possible to specify height and width, with the width being three times the height to be used for my LW sequence. Otherwise, I’m still interested in how this more abstract concept gets translated.
“What if outer space were udon” (CLIP guided diffusion did really well, this is cherry-picked though: https://twitter.com/nshepperd1/status/1479118002310180882)
“colourless green ideas sleep furiously”
This is a great example of how even a single iteration on the prompt can vastly improve the results.
Here are the results when using your quotes exactly:
Pretty dreadful! But here they are, with the exact same prompt, except with ”, digital art” appended to it:
Hmm.. the second iteration of the second prompt isn’t able to be pasted in that comment for some reason. Here it is:
Mr. Bean makes a mistake. (or charlie chaplin)
Far side comic by Gary Larson.
Overlapping mirrors.
Mona Lisa parody.
Oldest House Threshold from Control
A painting of a steampunk city that looks like a mix of San Francisco, Chongqing, and Quya. (Optional: replace Quya with Saint Denis).
A beautiful painting that encapsulates the feeling of love
Bilbo meeting Gollum for the first time. or Harry Potter fighting against dementors.
The Fighting temeraire by Turner The Mona Lisa by Vinci. Judith by Caravaggio.
I’m curious to see if it is able to remember famous paintings.
Avian bird city in the style of Studio Ghibli
“The pain of existence”