There are many advances that have been experienced in the field of AI (artificial intelligence) in recent years. However, it seems that these keep the ability to surprise intact, as Dalle-2 has shown. This system, heir to Dalle-1, has generated a real revolution in networks. And what is the reason for such a stir? As well, Dalle-2 is a system capable of generating realistic images or illustrations from descriptions written by its users.
This is the main functionality and the one that has gained the most prominence, with many images shared by those who have been able to test this new system, but it is not the only one.
Dalle-2 allows you to create images, edit them and version them
The artistic possibilities are multiplied thanks to Dalle-2. You will be able to obtain designs that respond to your ideas just by writing them, but also by editing existing images. The ability to add new objects or elements, as well as replace one with another easilyIt falls within the range of opportunities. All this, while the program takes into account the lights and shadows, the textures and the reflections of the original image, generating a very complete editing experience.
The third, and last, feature of this AI is the creation of new designs that reinterpret existing onesbeing able to vary aspects such as angles and perspectives, among others.
Dalle-2, the second parts are good
Before Dalle-2, of course, there was Dalle-1. This system was launched in January 2021 by the artificial intelligence research and development company OpenAI, and its aims were very similar to those of the current version: automatically generate images from descriptive texts. As can be seen in the image, the AI achieved its purpose, although the final result could not be considered a work of art.
A year later, in January 2022, after various updates, Dalle-2 is presented. Among its improvements, the achievement of a higher resolution and image quality stands out, as well as a better understanding and execution capacity. To this we must add the two features mentioned above: editing and versioning.
Within the reach of a few
At the moment, access to this AI is only in the hands of a select few, who are sharing Dalle-2’s creations with the rest of the mortals in their networks. In order to be part of such a select group, you can access a form through . Once this is done, they go on the waiting list until access is granted or not.
How does Dalle-2 work?
This AI works like a complex neural system, in which various images and their corresponding descriptive texts are intertwined. But the intricate technology of Dalle-2 (hereinafter referred to as our neuron friends) has evolved beyond simple connections.
These neurons have evolved in such a way that they are now capable of relating not only the subject (for example «teddy bears»), but also the action («mixing sparkling chemicals as mad scientist») and the artistic style («as digital art”). But these neurons do not offer us a single option, but they are capable of generating multiple proposals that respond to the description made.
Although this system is still under development, the qualitative leap has been most remarkable, which leads us to imagine a whole world of possibilities. Currently, Dalle-2 can only create faithful reproductions according to known parameters.. That is, his neurons are still soaking up knowledge, and what he does not know, he interprets.
What will be the impact of Dalle-2
Like all technological advances, or of any kind, this one has advantages and disadvantages. How can the use of this type of system affect ordinary people in the near future? And, above all, how will it affect design and graphic arts professionals? The disruptive that Dalle-2 raises is complex, encompasses multiple issues, from the eternal human versus machine capabilities debateeven topics such as copyright and copyright.
Predecessors
It is not the first time that OpenAI has developed an API capable of performing tasks that are considered exclusive to human beings, the GPT is a clear example of this. Like the Dalle, it has several versions, being the most recent. This system was launched in 2020 and its function is complete the text tasks that the user requests. The possibilities that this AI poses range from the creation of html codes to the creation of articles from already started sentences.
This is all because, like the neurons in Dalle-2, GPT-3 has been soaked in a monumental amount of information. More than 400,000 million texts make up the great conceptual map that GPT-3 has in its system, which can lead to consider the dangers of a tool capable of generating huge amounts of text in a very short time. Texts with a mostly guaranteed meaning and coherence, but the same does not happen with their veracity. In fact, its own creators came to withdraw its second version considering it a potential threat, capable of generating misinformation, of being used inappropriately.
Although, like GPT-3, Dalle-2 it is not a perfect system. And even so, it could become an opportunity for those whose artistic skills are not as developed, allowing them to obtain graphic representations of their ideas. It can also be useful for creating sketches before developing more complex designs or as a shortcut to test how an idea fits into different styles or trends. Likewise, it could become a great source of inspiration for design professionals, enriching their artistic background. For the time being, we can only wait with anticipation for the results of this advance and continue to let ourselves be surprised by the course that new technologies are taking.
Image:
Stay informed of the most relevant news on our news channel
Facebook Comments
Disqus