Google Imagen AI image generator

[The Mass Shooter Ron Soye]

Google has announced a new text-to-image generator called Imagen, similar to OpenAI's DALL-E 2. Supposedly, it outperforms DALL-E 2.

Just like OpenAI, you can't get access to the fun toy because of DANGEROUS DISINFORMATION. And if this was in our dirty hands we would use it to troll immediately.

But enough about that, let's see how good it is.

An art gallery displaying Monet paintings. The art gallery is flooded. Robots are going around the art gallery using paddle boards.
A dragon fruit wearing karate belt in the snow.
Sprouts in the shape of text 'Imagen' coming out of a fairytale book.
A majestic oil painting of a raccoon Queen wearing red French royal gown. The painting is hanging on an ornate wall decorated with wallpaper.
A Pomeranian is sitting on the Kings throne wearing a crown. Two tiger soldiers are standing next to the throne.
The Toronto skyline with Google brain logo written in fireworks.
A photo of a Corgi dog riding a bike in Times Square. It is wearing sunglasses and a beach hat.
A marble statue of a Koala DJ in front of a marble statue of a turntable. The Koala has wearing large marble headphones.
A transparent sculpture of a duck made out of glass. The sculpture is in front of a painting of a landscape.
A single beam of light enter the room from the ceiling. The beam of light is illuminating an easel. On the easel there is a Rembrandt painting of a raccoon.
A wall in a royal castle. There are two paintings on the wall. The one on the left a detailed oil painting of the royal raccoon king. The one on the right a detailed oil painting of the royal raccoon queen.
A blue jay standing on a large basket of rainbow macarons.
A bucket bag made of blue suede. The bag is decorated with intricate golden paisley patterns. The handle of the bag is made of rubies and pearls.
Three spheres made of glass falling into ocean. Water is splashing. Sun is setting.
A cute corgi lives in a house made out of sushi.
A giant cobra snake on a farm. The snake is made out of corn.
A bald eagle made of chocolate powder, mango, and whipped cream.
An alien octopus floats through a portal reading a newspaper.
A strawberry mug filled with white sesame seeds. The mug is floating in a dark chocolate sea.
A brain riding a rocketship heading towards the moon.
A small cactus wearing a straw hat and neon sunglasses in the Sahara desert.
Android Mascot made from bamboo.
A robot couple fine dining with Eiffel Tower in the background.
A photo of a raccoon wearing an astronaut helmet, looking out of the window at night.
An extremely angry bird.
A dog looking curiously in the mirror, seeing a cat.
A chrome-plated duck with a golden beak arguing with an angry turtle in a forest.
Teddy bears swimming at the Olympics 400m Butterfly event.

You'll notice that they created a lot of animal-themed pics. Maybe they did this because they're more shareable, or the generator is better at creating them, or legal concerns about the generator pulling in random photos of people. They didn't want to be raycis.

In the image of the duck arguing with the turtle, the turtle has a golden shell but probably shouldn't. The teddy bear image only has a single teddy bear.

I would like to see more comparisons like these from DALL-E, because the ability to get infinite variations on the same prompt is what will really kill the starving artists:

View attachment 3307388

View attachment 3307390

Related threads:

OpenAI’s DALL-E AI image generator can now edit pictures, too - Researchers can sign up to preview it
NeuralBlender - Create shitty (or nice) warped reality chaos.
AI-art isn't art - DALL-E and other AI artists offer only the imitation of art

https://wccftech.com/google-imagen-is-a-new-ai-driven-image-generator-capable-of-creating-images-from-words/

https://siliconangle.com/2022/05/24/google-details-new-cutting-edge-image-generation-ai/

 

[King Fructose]

I sure hope this doesn’t get abandoned tomorrow.

 

[comfysocks]

Stop hoarding the ai and let randos play with it. I wonder what kind of trolling got them privated.

 

[The Mass Shooter Ron Soye]

Stop hoarding the ai and let randos play with it. I wonder what kind of trolling got them privated.

OpenAI and Musk have been screaming the "friendly AI" mantra, disinformation and deepfakes are the new boogeyman, and we all remember Microsoft's Tay.ai.

 

[Jimmy Olsen]

I'm sure the government won't find pernicious uses for this technology.

 

[The Last Stand]

The dog sushi one is cute, NGL.

 

[Slav Power]

wtf this is AI generated?

 

[The Mass Shooter Ron Soye]

wtf this is AI generated?

They haven't made it possible to independently confirm that, but yes. You type the text on the left, you get an image like the one on the right.

If they typed "A painting of a human-sized banana relaxing on a sofa|recamier" or "draw a banana like one of your French girls", it might create an image that looks like your avatar.

 

[The Last Stand]

They haven't made it possible to independently confirm that, but yes. You type the text on the left, you get an image like the one on the right.

If they typed "A painting of a banana relaxing on a sofa" or "draw a banana like one of your French girls", it might create an image that looks like your avatar.

When is this going to be public? I want to have fun with this technology.

 

[The Mass Shooter Ron Soye]

When is this going to be public? I want to have fun with this technology.

There are several ethical challenges facing text-to-image research broadly. We offer a more detailed exploration of these challenges in our paper and offer a summarized version here. First, downstream applications of text-to-image models are varied and may impact society in complex ways. The potential risks of misuse raise concerns regarding responsible open-sourcing of code and demos. At this time we have decided not to release code or a public demo. In future work we will explore a framework for responsible externalization that balances the value of external auditing with the risks of unrestricted open-access. Second, the data requirements of text-to-image models have led researchers to rely heavily on large, mostly uncurated, web-scraped datasets. While this approach has enabled rapid algorithmic advances in recent years, datasets of this nature often reflect social stereotypes, oppressive viewpoints, and derogatory, or otherwise harmful, associations to marginalized identity groups. While a subset of our training data was filtered to removed noise and undesirable content, such as pornographic imagery and toxic language, we also utilized LAION-400M dataset which is known to contain a wide range of inappropriate content including pornographic imagery, racist slurs, and harmful social stereotypes. Imagen relies on text encoders trained on uncurated web-scale data, and thus inherits the social biases and limitations of large language models. As such, there is a risk that Imagen has encoded harmful stereotypes and representations, which guides our decision to not release Imagen for public use without further safeguards in place.

Finally, while there has been extensive work auditing image-to-text and image labeling models for forms of social bias, there has been comparatively less work on social bias evaluation methods for text-to-image models. A conceptual vocabulary around potential harms of text-to-image models and established metrics of evaluation are an essential component of establishing responsible model release practices. While we leave an in-depth empirical analysis of social and cultural biases to future work, our small scale internal assessments reveal several limitations that guide our decision not to release our model at this time. Imagen, may run into danger of dropping modes of the data distribution, which may further compound the social consequence of dataset bias. Imagen exhibits serious limitations when generating images depicting people. Our human evaluations found Imagen obtains significantly higher preference rates when evaluated on images that do not portray people, indicating a degradation in image fidelity. Preliminary assessment also suggests Imagen encodes several social biases and stereotypes, including an overall bias towards generating images of people with lighter skin tones and a tendency for images portraying different professions to align with Western gender stereotypes. Finally, even when we focus generations away from people, our preliminary analysis indicates Imagen encodes a range of social and cultural biases when generating images of activities, events, and objects. We aim to make progress on several of these open challenges and limitations in future work.

The best of these algorithms so far are all created by companies or institutions with similar stances on making their work available.

 

[The Last Stand]

The best of these algorithms so far are all created by companies or institutions with similar stances on making their work available.

Oh wow, me asking AI to make a pizza on a table is dangerous. GTFOH.

 

[The Mass Shooter Ron Soye]

Oh wow, me asking AI to make a pizza on a table is dangerous. GTFOH.

If the gloves come off, you could use a tool like this to generate infinite amounts of pornography of all kinds, images of any two celebrities or politicians boxing each other, anything needed to illustrate a fake news event you wrote up, etc. You could create Ben Garrison style cartoons in seconds, as seen in the Imagen sprouts example you would just need to put the text in quotes to insert it, and describe each panel.

Your imagination is the limit, and your shitty artistic abilities and laziness no longer shackle you. Only these AI overlord gatekeepers shackle you now.

 

[The Last Stand]

If the gloves come off, you could use a tool like this to generate infinite amounts of pornography of all kinds, images of any two celebrities or politicians boxing each other, anything needed to illustrate a fake news event you wrote up, etc. You could create Ben Garrison style cartoons in seconds, as seen in the Imagen sprouts example you would just need to put the text in quotes to insert it, and describe each panel.

Your imagination is the limit, and your shitty artistic abilities and laziness no longer shackle you. Only these AI overlord gatekeepers shackle you now.

THEY made it, now they're afraid of what WOULD happen? I mean, THEY MADE IT.

 

[Syntaxion]

It will just be 45% porn, 45% edgy/offensive things and 10% actually interesting normal stuff

 

[Erich Raeder]

I'm going to be honest this is giving me a real inferiority complex. If this works as demonstrated I'm fully 100% certain I would never ever be able create something like this no matter how much effort I would put into it. The people working on this are on another level. What a reality check.

@Null you might find this interesting, ties a bit into the dead internet thing you talked about in the last MATI stream and deepfakes you also mentioned once or twice.

 

[Manul Otocolobus]

"A kiwi bird yelling "Nigger!" at the top of its lungs!"

 

[Maurice Caine]

If the gloves come off, you could use a tool like this to generate infinite amounts of pornography of all kinds, images of any two celebrities or politicians boxing each other, anything needed to illustrate a fake news event you wrote up, etc. You could create Ben Garrison style cartoons in seconds, as seen in the Imagen sprouts example you would just need to put the text in quotes to insert it, and describe each panel.

Your imagination is the limit, and your shitty artistic abilities and laziness no longer shackle you. Only these AI overlord gatekeepers shackle you now.

Eh if Google did it then I figure by the end of the decade we'll have something similar to the masses. Can't hoard shit forever.

 

[PuffyGroundCloud]

I'm really impressed with the result
I wonder if the images look good because it's in a "controlled" data environment

 

[The Mass Shooter Ron Soye]

Eh if Google did it then I figure by the end of the decade we'll have something similar to the masses. Can't hoard shit forever.

Yes, they are delaying the inevitable. Hopefully some computer scientist assholes will pool their RTX 3090s together and train models that are about as good. Or we could get leaks out of the companies.

I'm really impressed with the result
I wonder if the images look good because it's in a "controlled" data environment

The bald eagle one impressed me. It looks like it invented a legit cute dessert item. The strawberry mug is particularly cursed.

There's no way to tell how much they curated it, but you can find some subtle mistakes.

 

[Saint Agustin]

Can't wait for tranny suicide AI generated photos.