OVERRIDE PROJECT DEEPDREAM V2.0 PROTOCOLS///

10年くらい前のAIの絵覚えてる？ 『DeepDream』は衝撃の気持ち悪さだったよね。

also before transformers/attention, even with basic GAN's it's still basically divining functions between noise and your training data, and the model builds a feature map. One of the first viral ML image gen experiments deepdream was created by extracting features from a CV model

Which generative AI? A word n-gram language model? Deepdream? I'm sure those gave as good results as the ones we have today! The AI boom started a decade ago! Not everyone who stumbled upon social media has a lawyer, that was kind of the charm of social media, to be for everyone

That seems clearly wrong? Surely adjusting x only amount to forming an example period (e.g. DeepDream)?

Who the F cares‽ It is the collection of clips edited together and the ability to just enjoy the “entire” clip for what is intended, “humor” that is what is germane! And I do not GAF if it was created with: • AI •Stable Video Diffusion •OpenAI Sora •Luma Dream Machine •Kaiber •Pika •Runway Gen-2 •Midjourney •Deforum •Stable Diffusion •EbSynth •Disco Diffusion •OpenAI DALL·E •OpenAI CLIP •D-ID •Reface •Synthesia •DeepFaceLab •Runway ML •StyleGAN •GAN-based Face Reenactment •DeepFake Autoencoders •IBM Watson Trailer AI •LSTM Script Generators •DeepDream •DALL·E •Midjourney •Stable Diffusion •Adobe Firefly •Runway •Processing •p5.js •TouchDesigner •Max/MSP/Jitter •Blender •Adobe Photoshop •Adobe Illustrator •CorelDRAW •Painter •Procreate •MacPaint •Deluxe Paint •Paintbrush •Amiga HAM graphics modes •Sketchpad •Pen plotter art (HP, CalComp) •Oscilloscope vector art •Early CAD systems adapted for art •ASCII art •FORTRAN /BASIC/ALGOL graphics routines •Line-printer art •Fractals (Mandelbrot sets) •L-systems •Cellular automata I just want to laugh, cry, be shocked or inspired, be amazed at technique, reminisce, and enjoy creativity. What I do not want is self-appointed “𝕏 POLICE” chest thumpers, proclaiming every artisticly created, co- mingled post in the negative, as if it matters for the reason I like them. So please, stay out of feeds with that illiterate, inexperienced nonsense and let those that appreciate them, appreciate them! (And YES! I have used them all of the above! I started my “computing career” with the following…

#aicreations #aiart #DeepDream

Les artefacts des CNN type Deepdream nous rappelaient notre manière de rêver mais n'étaient que des hallucinations ?

Two ish years ago when AiArt got on the public GroupThink vs being DeepDream Corgi/Eyes nightmare fuel Image comics released a statement they'd NEVER use AI or let anyone who did work for them. Nice "Virtue Signal" but it seemed a month after that I saw this new Image comic on the shelf - with this cover which is obviously Midjourney V3 era style and probably the face worked on a bit. My issue is that for all the "Yaa Ai Slop!" anonymous trolls have screamed we've had a decade of bland "Corporate SLOP" via a handful of giant companies including eating their young like they did the 80s Direct market and we just get the re-re-re-re-relaunch of "Bat-Mutant"... Especially it's usually hard to tell one illustrator from another due to PhotoSLOP aka they all Google images and TRACE them, competing for a race to the bottom with 3rd world illustrators. My favorite is Greg Land who started in Crossgen but does Marvel PhotoSlop and is so bitter he blatantly traces Pronography... His women change body in the same issue and it seems the only whine is they look S- ual and "Teen" characters look like late 20s jacked up pron stars in weird twisted poses. I hope they don't order him to draw Power Pack without being able to age the characters up or the man will end up jailed. (or use Clip Studio Paint with 3D props!) So this "Ai" stuff is mostly "HAllp us! Turkin R Jurrbz!" for what - the 1/2 issue of the re-re-re-re-re launch of Bat-Mutant, competing with 3rd world hacks for Google tracing some random image for $30? The big companies might as well use it and train it on THEIR IP. I'm for use of Ai - I like the STORY and if there's a way to bypass increasingly hard to deal with "Illustrators" who mostly only can TRACE images... However - pic 3 and 4 - I'm ALL FOR a "Human Only League" BUT we need to DITCH the computer save the last 30% of correcting minor errors and preparing it for publication. The last bit cost a LOT of "Jurrbz" btw.... BUT the comics need to be drawn by HAND on PAPER. Or canvas. NO TRACING and that means a light table (including over a monitor) and light projection or camera obscura/lucida. They can use any art reference including AI but it's EYE to MIND to HAND to PAPER. Original art and stories, creator owned. Some kind of Consignment network to push boxes of random comics around comic shops and IF they sell the store gets some of the $. This would be a re-launch of the Underground Comix movement and/or early Direct Market. NO web distribution and crowdsource does NOT count - the web "Major Search Engine" is biased so you won't find the new stuff until years AFTER - even if you know 100% what you are looking for you wade through 10 pages of spam and scams. We need random crazy indies IN THE STORES on a consignment thing. Stores should love "Free Inventory" especially those Coolsville style new generation comic stores that seem to have so much empty space with ONLY Marvel, DC/Image and fake "Indies" hidden behind the PILE of Funko...

How to train a neural net. Objective, gradients, backprop, DAGs, debugging, deepdream.

i knew it was going to blow up since deepdream