I don’t really know, but I think it’s mostly to do with pentagons being under-represented in the world in general. That and the specific way that a pentagon breaks symmetry. But it’s not completely impossible to get em to make one. After a lot of futzing around, o1 wrote this prompt, which seems to work 50% of the time with FLUX [pro]:
An illustration of a regular pentagon shape: a flat, two-dimensional geometric figure with five equal straight sides and five equal angles, drawn with black lines on a white background, centered in the image.
Fun Fact: It is very difficult to get any of the image generators to make a pentagon.
To choose order over disorder, or disorder over order, is to accept a trip composed of both the creative and the destructive. But to choose the creative over the destructive is an all-creative trip composed of both order and disorder. To accomplish this, one need only accept creative disorder along with, and equal to, creative order, and also willing to reject destructive order as an undesirable equal to destructive disorder.
The Curse of Greyface included the division of life into order/disorder as the essential positive/negative polarity, instead of building a game foundation with creative/destructive as the essential positive/negative. He has thereby caused man to endure the destructive aspects of order and has prevented man from effectively participating in the creative uses of disorder. Civilization reflects this unfortunate division.
THE CURSE OF GREYFACE AND THE INTRODUCTION OF NEGATIVISM - Principia Discordia
He’s Weyland.
an eight-year-old girl was among those killed
Hop on Adobe stock right now and search for something. Half of the results will be AI-generated. There’s a search filter that can exclude them.
As a human bot in b2b marketing, the dark arts are antithetical to learning. To market is to forget.
I think it was Perplexity. Moved to using Flux since that cuddly monstrosity.
I kinda assume the internet is just the same 200 dudes from Sensible Erection migrating from platform to platform.
Is it Elmo?
Calling what attention transformers do memorization is wildly inaccurate.
*Unless we’re talking about semantic memory.
It honestly blows my mind that people look at a neutral network that’s even capable of recreating short works it was trained on without having access to that text during generation… and choose to focus on IP law.
The issue is that next to the transformed output, the not-transformed input is being in use in a commercial product.
Are you only talking about the word repetition glitch?
How do you imagine those works are used?
It’s called learning, and I wish people did more of it.
This is an inaccurate understanding of what’s going on. Under the hood is a neutral network with weights and biases, not a database of copyrighted work. That neutral network was trained on a HEAVILY filtered training set (as mentioned above, 45 terabytes was reduced to 570 GB for GPT3). Getting it to bug out and generate full sections of training data from its neutral network is a fun parlor trick, but you’re not going to use it to pirate a book. People do that the old fashioned way by just adding type:pdf to their common web search.