Launched in early 2021 by the company OpenAI, founded by the renowned Elon Musk, the artificial intelligence called DALL-E (a fusion of the famous painter Salvador Dali and Disney’s robot AI character Wall-E) made a big splash not only in the tech industry but also in the worlds of designers, graphic artists, and various other domains involving image creation and manipulation.
DALL-E is, in fact, an image generator that operates based on textual descriptions, building upon a previous algorithm that garnered significant media attention, GPT-3. While the results weren’t initially of exceptional quality with the initial version, they have seen significant improvement with its latest iteration, DALL-E 2.
Without further delay, we were eager to try out this new tool and form our own opinion about OpenAI’s latest flagship. To do so, one needs to sign up on the official website, which is now open without a waiting list. OpenAI’s initial test for us was to attempt to recreate the DataScientest logo using a simple textual description.
So, after a few attempts where we learned to fine-tune our queries, the description “a logotype of a cyan winged lion, seen in profile, sitting with a crown on its head” allowed us to obtain the following result in just a matter of seconds:
or
compared to our original logo:
It’s worth noting that the small colored squares in the lower right corner of the image serve as a trademark of the algorithm, ensuring the image’s origin.
Furthermore, it’s essential to recall that DALL-E is not aware of specific entities and proper names, such as public figures, companies, or any other private rights, to prevent its tool from generating violent, sexual, or inappropriate content. Therefore, it was impossible for us to simply type “DataScientest logo” to obtain the best result, nor could we establish a direct link with our company. DALL-E lacks this knowledge.
Given these results, we can be more than satisfied, as the resemblance is striking at first glance, and the details are correctly preserved.
Lastly, this artificial intelligence also offers additional features, such as image editing or integrating elements into an existing image, among others, but we’ll let you explore these features for yourself!
But what is the impact of such a creativity amplifier?
As this technology is no longer the sole property of a single company with massive resources, we can confidently bet on even more widespread and accessible democratization than currently exists. Recent releases of open-source algorithms, such as crAIyon (the new name for DALL-E mini), Stable Diffusion, or MidJourney, indicate this trend. But should professional illustrators and other image creators be concerned about the democratization of this kind of technology? Not really.
In fact, for many of them, this technology should be used as a supplement rather than a replacement, especially in the preliminary phases of a project. Serving as the initial sketch of a work, DALL-E can inspire these artists without worrying about intellectual property, as the generated images belong entirely to us and can be reworked later by true professionals.
While DALL-E’s output is impressive, it is less effective in handling details, backgrounds, and the overall qualities of an image that make it professional work.
This tool can be used to seek ideas, put an image to words, explore an initial concept without investing a day’s work, and it’s a great assistant for those who suffer from creative blocks.
In summary, the artistic creation process will likely evolve alongside artificial intelligence, just as it did with the arrival of now-ubiquitous software like Photoshop or InDesign, which already incorporate AI-based tools.
This is precisely what Microsoft is attempting with the recent release of Microsoft Designer, a new content creation application aimed at competing with Canva. It integrates the new artificial intelligence DALL-E.
While its usage and access are somewhat similar to the DALL-E website, the integration of this tool into Microsoft’s software indicates significant interest from major tech players in this AI. It confirms the growing accessibility of the tool across major platforms.
But in this case, shouldn’t we be concerned about homogenization in the creative process? Isn’t DALL-E an open door to the spread of deep fakes?
If you want to understand how DALL-E was developed and how this AI works, we invite you to enroll in our deep learning training, available here.
