AI is taking Twitter by storm – create works of art from words for you too

Technology is constantly evolving and with it the creativity of artificial intelligence. Now there are even AIs that can turn written text into really cool or funny images, winning over Twitter.

What AI is it about? There are some AIs capable of creating images from written text. Two of the best known are Dall-E from openai.com and Midjourney (via midjourney.com). The third AI in the group is Craiyon, for which a separate Twitter account was created.

Each of these three AIs is able to create an image from a written text. Some of these pictures look really good, others are just hilarious. So does this very special Pikachu:

Recommended Editorial Content

At this point you will find external content from Twitter that complements the article.

I consent to external content being displayed to me. Personal data can be transmitted to third-party platforms. Read more about our privacy policy.

AI artist as Discord bot

How can you let the AI ​​draw for you? Both Dall-E and Midjourney require you to request access to the beta test on the company’s website. This access will be granted in waves, so it may take some time before you know if you can test the AI.

If you get access – which you will be informed by email – creating images using AI is very easy. You enter a few words in the text field, wait a few minutes while the AI ​​creates your image and you have your own work of art. Midjourney even offers a Discord bot that you can integrate on your Discord server.

If you don’t get access to Dall-E or Midjourney, Craiyon is still available for free at craiyon.com. Craiyon was previously known as Dall-E mini but changed the name at OpenAI’s request to avoid confusion with Dall-E.

Microsoft is working on new AI that will make raid bosses invincible in MMOs

From demogorgons to mushroom cities in snow globes

How does social media react? The internet is excited about the works of the AIs. There’s even an extra account on Twitter called Weird Dall-E Mini Generations (@weirddalle via Twitter) with over a million followers who posts crazy creations from Dall-E mini or Craiyon.

Weird Dall-E Mini Generations’ various images and collections regularly receive many thousands of likes and retweets.

For example, the account’s pinned tweet shows Demogorgon, known from the hit series Stranger Things, with a basketball and received over 160,000 likes and was shared over 13,000 times.

Recommended Editorial Content

At this point you will find external content from Twitter that complements the article.

I consent to external content being displayed to me. Personal data can be transmitted to third-party platforms. Read more about our privacy policy.

Craiyon’s images of a Demogorgo holding a basketball

But not only Craiyon creates fascinating pictures that are shared on social media. For example, Twitter user @fabianstelzer showed an impressive comparison between Dall-E and Midjourney.

He fed both AIs the words “A mushroom city in a snow globe sphere, stunning detail, hyperreal rendering” and shared the different images.

Recommended Editorial Content

At this point you will find external content from Twitter that complements the article.

I consent to external content being displayed to me. Personal data can be transmitted to third-party platforms. Read more about our privacy policy.

The mushroom town in a snow globe: Dall-E on the left, Midjourney on the right

AI creates images from memory

How does the AI ​​work? Boris Dayma and Pedro Cuenca explain how the Craiyon AI works on the Weigts & Biases website (via wandb.ai). Boris Dayma trained the AI ​​according to craiyon.com. Pedro Cuenca worked on the backend. Regarding how Craiyon works they say:

  • The AI ​​would be trained by looking at millions of images from around the web with their accompanying captions.
  • The AI ​​learns over time how to draw an image based on a text prompt. In doing so, the AI ​​learns some concepts from memory as it may have seen similar images.
  • The AI ​​can also learn to create unique images that don’t exist yet. As an example, they cite an image that goes with the text “the Eiffel Tower lands on the moon”. To do this, the AI ​​combines several concepts. In order to do that, you have to combine several models. These models include:
    • an image encoder that converts raw images into a sequence of numbers, with the associated decoder
    • a model that converts text input into an encoded image
    • a model that judges the quality of the generated images for better filtering

AI can play Minecraft like a human – only had to watch 70,000 hours on YouTube

What is all this good for? The question that remains, of course, is: What is it all good for? In addition to creating funny or interesting images, there must also be a practical use.

How the AI ​​will work in the distant future and what people will use it for is of course still pure speculation, but functions such as drawing identikit images are quite conceivable.

What do you say about the AI ​​that can draw you a picture? Do you think this is an interesting and potentially even useful technology, or do you say it’s just an unnecessary gimmick? Let us know in the comments!

In another project, researchers wanted to prove what evil AI could do. You can see the result of the investigation on MeinMMO.

Researchers want to prove what an evil AI can do – It designs 40,000 chemical weapons in 6 hours


#Twitter #storm #create #works #art #words

Leave a Comment

Your email address will not be published.