google.com, pub-9309777685902732, DIRECT, f08c47fec0942fa0 google.com, pub-9309777685902732, DIRECT, f08c47fec0942fa0
top of page
Writer's pictureRiva Kajangu

Text-To-Image Is Here, Hands-On With The DALL-E AI Tool - (Original Post 08.22.22)


I love technology, and testing it firsthand is the ultimate experience. Here is what DALL-E generated for me.




I am sure you’ve read or heard about the text-to-image artificial intelligence tool coming out; if you have not, here is a little taste of it.

There are a few companies and projects behind these tools. I tested DALL-E, “a new AI system that can create realistic images and art from a description in natural language.” It's built by OpenAI.

We’ve trained a neural network called DALL·E that creates images from text captions for a wide range of concepts expressible in natural language.

It’s pretty straightforward once you get access to use it.

My experience

First, I received the email with the invitation (pic below).



Picture of Authors invitation The message made me feel special; although I know not everyone will feel the same, this is my experience.

As a technologist, I know we are at the forefront of the next tech evolution; some are calling it web 3.0. The best explanation I heard about this next wave was from Eric Schmidt, the former CEO of Google, who said web 3.0 is about ownership of content (paraphrasing).

For a brief overview, at the beginning of the internet, people were only able to read stuff online. The next wave allowed people to read and write, and now owning is here.

After clicking on “Get started,” a rush of excitement entered my body. I was about to challenge a machine to create incredible images and challenge myself to come up with crazy ideas to test how intelligent the system was.




At this point, I’ve seen some images created and knew the potential of DALL-E’s possibilities. Then an empty text box greets you (pic below). Now I am frozen for a second. Now the thinking starts, what in the world do I tell it to create for me?

At the bottom of the page, they give you a few examples to get inspiration from, but first, you feel that you cannot come up with anything close to those samples.

Likely for me, I already had an idea of what I wanted to see. I am writing a fictional story about aliens, the people God created before humans called Luminarians, who are tucked away somewhere in the universe. I have a description of them, and I asked DALL-E to visualize it for me, and here it is. Here are some other requests I made:

 

Adam and Eve in a marketplace - Dalle-2 Image
Adam and Eve in a marketplace

a strong baby doing push ups in a crib - Dall-E 2
a strong baby doing push ups in a crib

The cookie monster eating an apple in a grass field - Dalle-2 Image
The cookie monster eating an apple in a grass field

A genius boy playing on top of a tree- Dall-E 2
A genius boy playing on top of a tree

A chemist person on a dance floor with a disco ball - Dalle-2 Image
A chemist person on a dance floor with a disco ball

A realistic alien flying on a spaceship to earth - Dalle-2 Image
A realistic alien flying on a spaceship to earth

A dog walking a tiger in a dream - Dalle-2 Image
A dog walking a tiger in a dream

 

Overall the images depict precisely what you describe, and it does an amazing job filling in the environment to give you more than a character just standing on a blank background.

There are quite a few reviews about the AI tool, but I did not want to read them all till I could test it first and give it a fresh perspective. I know there are efforts to deal with bias, negative depictions, and other considerations.




Here is my take thus far, which is only with 20 generated images.


1. Some results were not diverse without giving it a specific race type.


2. With the two examples of a baby and a boy, it's clear those are white children. I have questions if the tool will only display a fair skin person when no specific description is given for a scene, like a strong baby or genius boy.


3. I wonder who the tool associates smart adjectives like; agile, alert, brainy, bright, brilliant, clever, intelligent, and keen the most.

4. Now it always gives you four results; some images include four different types of people, like the first image below. and some not so much, the second image. I have a designer friend who is a bit worried that a tool like DALL-E will replace him. I told him that he could become a better artist because he already has an excellent eye for creating stuff, which the AI tool will only help him do better, faster and cheaper.

In conclusion

I believe technology will help us make the world a better place to explore and discover. We cannot deny that bad actors are always around and will try to create chaos, but most of us will keep moving forward in imagining a better future for us all.

DALL-E and the other text-to-image tools will improve our visual world and maybe birth the next Picasso or Van Gogh of our generation. When knowledge and tool are democratic, we can surprise ourselves as human beings. We are all creative in some capacity.

I will continue to test the AI tool and see how far I can push my creativity, and the AI will continue to learn from us and improve itself over time.


 

Follow me for more, and I’ll take you on a journey. https://medium.com/@thekajangu

Comentarios


bottom of page