A few months ago we commented on the keys to Artificial Intelligence for the year 2022. Customer service, problem prevention and making the most of 5G were the most highlighted phrases in bold in that post.
Some time later, the networks have exploded with the discovery of a new AI operation. It is DALL-E Mini, an online tool that draws anything you ask it to. In this Befree blog post we take a closer look at DALL-E and see what it has to offer.
What is DALL-E Mini
This artificial intelligence functionality basically consists of drawing whatever you ask it to draw. It is a generator from the company OpenAI – who we will talk about later – created by Boris Dayma, who you can follow on Twitter to keep up to date with the latest news and updates on DALL-E Mini.
The tool analyses the words in the phrase you have asked for, searches for them on the Internet and finds images associated with them. Finally, DALL-E Mini creates a composition by joining fragments of these images and tries to make the result as realistic as possible.
Walter White, from the series ‘Breaking Bad’, drawn in the Van Gogh style by DALL-E Mini
The results, despite the good example shown above, are not always accurate. The AI sometimes produces completely bizarre and surreal images that can only arouse laughter in the person who suggested them.
However, as Xataka points out, all artificial intelligences go through a training period. After that, DALL-E Mini will have improved by leaps and bounds, so outlandish results are necessary evils for its refinement.
Where does DALL-E Mini come from?
The origin of this AI functionality lies in the battle between DALL-E 2 and a Google project called IMAGEN. In early 2021, the artificial intelligence research and development company OpenAI created DALL-E. It worked as an AI that drew whatever was asked of it by associating visual and textual concepts. In the course of 2022, OpenAI improved it with DALL-E 2.
Google did not want to be left behind and launched its own project: IMAGEN. According to its research, this AI obtained more accurate results than DALL-E 2. However, it is impossible for users to access IMAGEN, as Google has argued that it is better for it to remain a non-commercial product for ethical reasons.
So, in its battle against Google for artistic AI dominance, the idea behind DALL-E Mini is to create an algorithm to generate images with a smaller infrastructure and fewer hardware resources.
How it can be used
To access this free demo, OpenAI has provided a web space where everyone can try out their tool. Click on this link to go to the Hugging Face website, where DALL-E Mini is hosted.
You don’t have to be a computer expert or technology scholar to try it out. Just type in a sentence you want the AI to draw and wait for the tool to do its job. The sheer number of users means that the waiting time can be very long at times, or that DALL-E Mini doesn’t even have the capacity to start drawing your request. But it’s worth a try at least once.
