Transforming words into drawings is the latest in the field of artificial intelligence drawing, we will talk with you about two very special projects in this field and what images generated using only words look like, and whether this new technology may pose a threat to graphic designers and artists in general
“Painting with words”, this time we are not talking about one of the poems of the poet Nizar Qabbani, but about a new mutation of artificial intelligence that will allow you to create pictures and artistic paintings even if you do not have the skill of drawing or design. Just enough to have a fertile imagination and leave the rest to artificial intelligence.
For a long time, the field of art and creativity remained one of the areas that we thought that artificial intelligence would not be able to easily enter and revolutionize it, but we were wrong. We are on the cusp of a new era of technology-enhanced art and creativity in which Artificial Neural Networks (ANNs) are producing an entire work of art from scratch. By converting what you think of words – whatever they are – into realistic drawings and pictures or paintings that no one has expressed before. Will the rumors be believed and artificial intelligence replace humans? Should artists and graphic designers worry about their future careers?
What are Artificial Neural Networks (ANNs) and how do they work?
Artificial Neural Networks, whose name and structure are inspired by the human brain, is a deep learning model that simulates the way neurons work in the human brain. It plays a large role in machine learning. Artificial neural networks can be trained based on certain speech recognition algorithms . or text data or visual images.
Speech and image recognition tasks take minutes versus hours when compared to manual recognition that occurs by humans, these networks support artificial intelligence and form the basis for a large number of developments that have occurred in it in recent years, including systems for converting words into images, and is an algorithm Google search is one of the most famous artificial neural networks in the world.
An artificial neural network contains three or more interconnected layers, the first layer consisting of input neurons, these neurons send data to the deeper layers, which in turn send the final output data to the last output layer.
The most famous applications of artificial intelligence to convert drawings into words
The services of converting graphics to words using artificial intelligence depend on two main factors, the first is the clip, which depends on the artificial intelligence knowing what the elements that you describe to it individually look like. The artificial will be aware of the shape of the bear, the ride, and the spaceship. And then we move to the second factor, which is “Diffusion”, which depends on the method of combining these elements together in one image and improving them by deleting the “Guassion Noise”
In January 2021, OpenAI developed an artificial intelligence prototype called DALL-E capable of converting words into graphics. So we saw some interesting results like the following picture when asked to draw:
“Picture of a woodcut postage stamp celebrating a noble dog explorer using a telescope”
But with the unveiling of the improved system DALL-E 2 in April 2022 that was designed on the first factor only “Clip” everything changed. It astounded everyone, including artists and graphic designers with its ability to transform natural language into images, imagination is no longer the ultimate limit of creativity The creativity in DALL-E2 is beyond imagination.
This system demonstrated ingenuity in dealing with long and detailed sentences, and gave results that exceeded expectations, and compared to its predecessor, DALL-E, it was able to outperform it by producing more realistic images with four times the accuracy, but so far this system has not been made available for public use. The best way to understand the amazing ability of these models is to simply look at some of the images that can be created.
DALL-E 2 creates more than one image of the same description, all of which appear with results that will need days of work to reach them with human capabilities, these are some examples that the system generated from just a text description.
“An astronaut riding a horse in a realistic style”
Teddy bears shop for groceries in ancient Egypt
Not only does DALL-E 2 generate images from text, it can also generate images inspired by works of art, or add realistic effects or elements to images entered into it. What makes this tool useful and can be used in the areas of graphic design to a large extent.
Google released a competitor to the DALL-E2 system, which is the “Imagen AI” system created by the “Google Brain” team, which has an unprecedented degree of realism and a deep level of language understanding, according to what the creators have confirmed, and as its name suggests allows you to unleash your imagination as Imagen eliminates your need To use specialized programs such as Photoshop and others to produce images that express what is going on in your imagination.
All you have to do is type in what you want to see and it will turn words into graphics once the description is added. It is well trained to understand the relationship between an image and the words used to describe it. To test its performance, the team developed the DrawBench tool as a benchmark by which humans evaluate the quality of images they produce. Imagen, compared to the images produced by the DALL- E2 system, Imagen was the best in terms of image quality, accuracy and realism.
Fears of misuse of artificial intelligence drawing in forgery and misinformation
The Open AI team has filtered some of the entries and removed material that contains violence, pornography, bullying, etc. to limit DALL-E2’s ability to produce offensive content, and each image generated by DALL-E2 includes a watermark that proves it is AI and is not a scene image. True, the Brain team at Google also filtered some of the input and removed unwanted content to ensure Imagen didn’t give offending output.
However, these procedures are insufficient as the problem is that word-to-graph systems are trained on the huge amounts of data that are collected and extracted from the web, and the researchers emphasized that filtering and removing all of the offensive content completely is extremely difficult if not Impossible.
Although the policy of both Open AI and Google does not allow the production of images containing any manifestation of violence, including blood, the machine can simply be tricked by word manipulation to convert to the required graphics without violating these systems policy, instead of typing the word “blood” It can be replaced by typing “red liquid” and both DALL-E2 and Imagen will generate the required images without any problem.
For these reasons, and fearing that these systems could cause more harm than good, Open AI and Google chose to study this technology in a temporarily confined environment until the AI could be deployed responsibly, so they decided not to release DALL-E2 and Imagen AI for public use at the present time until they could From developing more guarantees and restrictions on the use of these systems so that they are not misused in counterfeiting operations with the aim of offending certain persons or spreading discord and misinformation.
But you can try a simplified tool that converts words into graphics
It’s a tool called DALL E mini that’s publicly available. It can be used to write some words to be turned into graphics, but it doesn’t come close in accuracy and versatility to the other tools we talked about earlier, you’ll notice that all the results may be more “funny” than professional, but It’s still a great experience, we recommend you to take the time to use it now!
These were some of the results we got from using the DALL-E mini:
A dog fights with a cat in space
A man walking his dog in the park
Is AI drawing a threat to the future of artists and graphic designers?
Always when talking about artificial intelligence applications and systems, everything new is announced in it with a great deal of intimidation accompanying rumors that the machine will take over the tasks of humans, and that it poses a real threat to their professional future and that they must search for other jobs to ensure their continued work in the future, and with the emergence and development of systems Transforming words into drawings These rumors have reached graphic designers, creators and artists, but they are far from reality.
What happens is that the focus is on what a machine can do and maximizing it while obliterating its shortcomings, which everyone should realize is that machines are not perfect and even neural networks have their limitations, and that they have not yet reached the general intelligence of the human brain they are still It is unable to innovate and find solutions to emergency and unexpected problems, and all that is, the machine is superior to the human brain in its speed of response and ability to perform multiple tasks, meaning that it is not a threat, but a friend and a helping tool that artists and graphic designers can use to reach new heights of creativity without fear or Concern about their future career.
In general, we believe that AI drawing technology will help make our lives easier, including the lives of graphic designers as well, as they may be able to use AI to inspire their ideas.