WelcomeUser Guide
ToSPrivacyCanary
DonateBugsLicense

©2025 Poal.co

(post is archived)

[–] 11 pts (edited )

Don’t be fooled.

The AI doesn’t draw things from scratch.

It uses a technique called collage to assemble different images (with transparent backgrounds) together based on the keywords. Pretty much like using layers in photoshop.

The hardest part was to assemble a huge database of images.

does the same thing by picking images online based on keywords and blend them together. Results are less impressive (due to things in images not always having a transparent background)

[–] 1 pt

Pretty much all the AIs are trained with human creations and then they combine them in their opaque ways into something that looks like it was created, that fools some into thinking the AI is like us, a [[[fellow human]]]. It helps if the observer wants to believe.

[–] 0 pt

Here's some text from the official OPENAI website , they say that:

DALL·E 2 has learned the relationship between images and the text used to describe them. It uses a process called “diffusion,” which starts with a pattern of random dots and gradually alters that pattern towards an image when it recognizes specific aspects of that image.

How can this DALL-E take "ideas" from a database of stock photos, and through a text command, create images that are entirely unique? How does "it" understand?

[–] 4 pts

It has to have a reference.

An AI can’t create something out of nothing.

The drawings shown in the video look more like 3D scenes with filters.

[–] 2 pts (edited )

I tested DALL•E Mini. Results below in screen capture.

Try for yourself: https://huggingface.co/spaces/dalle-mini/dalle-mini

It’s the only site I could find with any version of the software. If you find a different/better one, please share.

[–] 0 pt

starts with a pattern of random dots and gradually alters that pattern towards an image when it recognizes specific aspects of that image.

That's pretty novel. It's like the person who says, "I don't know how to make it, but I knows it when I sees it."

[–] 0 pt

Be interesting to see what turns up in a reverse image search.

[–] 0 pt

I'm not sure we'll get results. Changing hue and resolution might confuse joggle algorithm.