Apps to generate images based on artificial intelligence and machine learning became particularly popular in 2022. For several weeks the trends in social media they drove not only the name of specific tools, but also the creations of users who experimented with them: from the most realistic and somewhat intuitive, to some merely speculative and abstract things.
There is no doubt, in addition to being in a rehearsal stage, it is also a phase of democratization for this type of digital instruments.
In case you don’t know them, I’ll explain how they work roughly. beyond the name of software (we’ll explore examples later), you have to enter a word, phrase or concept. In some cases select a graphic style, pictorial school, or artistic trend. And after a few seconds the computations will throw you an image generated specifically for you based on the data you provided.
However, beyond its specific operation (there is a totally subjective argument, so deal with it), what is most fascinating of all this is not the files generated, but the discussions and current implications that this type of program is arousing.
Questions that surround ethical dilemmas, data privacy, aesthetic, racial and in some cases even espionage. In this article we will briefly review the first four points and leave the last one for a more robust and detailed delivery at another time.
IT MAY INTEREST YOU: ‘Severe’ cost cuts saved Twitter, says Elon Musk
THIS PERSON DOESN’T EXIST
One of the first tools I remember hearing about for generating realistic images using software was This person doesn’t exist. His description reads as follows: “Generate a random human face in 1 click and download it!”.
And it’s as simple as that. The two images below have been generated thanks to this artificial intelligence. Images that are based on the StyleGAN algorithm of NVIDIAwhich takes more than 100 different factors for graphic production, such as the position of the face, the person’s personality, gender, hair style, among others.
I chose two portraits of infants on purpose, although the options available are quite wide and at least in this case I did not notice that there was discrimination based on skin color, age or gender. And I say that I chose the portraits on purpose because if they existed, if these children were real, there would have to be a permission signed by their parents or guardians to be able to use their image in this article.
Or you would have to resort to smoothing or erasing some gestures so as not to compromise your integrity and personal data.
From here, the principle of generating images changes the way we do things and how those digital objects relate to their own environments. Will there come a time when these types of processes are mostly regulated? We still don’t know for sure.
The crucial thing here is how easy it is to generate a realistic image that can be used for any purpose. Illustrate an article like this, where it is made explicit that they are people who do not exist. Or they could be used to falsify documents, impersonate other people, and we will not give more ideas about that. Here, as in almost any case, your imagination is the limit.
DEEP NOSTALGIA
In February 2021, the MyHeritage company launched an artificial intelligence called Deep Nostalgia that allows you to animate static photographs of people. To date, according to the site, more than 105 million animations have been made.
Here what the algorithm does is look for facial features and give them movement, so that it appears that they are making gestures. This, it should be noted, only works with a photo of a human face.
If you follow the correct steps, You can see one of the results by clicking here.
The controversy here is that MyHeritage originally released a video that suggests deep fakes.
And it is not an isolated opinion. Comments about it arose in various forums such as Reddit. The company states that to avoid this the tool will not include options where the mouth moves or simulates speech, so that Deep Nostalgia does not imitate any speech.
DALL-E
This app caused a stir since it was in the testing phase, as well as with its respective versions. And now that it is in full version it is available to everyone.
The operation is simple. You introduce a concept, click on draw and the tool does the rest.
In this case, the results drew attention for how hilarious or absurd they could be. Here some examples.
Until September 2022, the company confirmed having interacted with 1.5 million users and generated more than 2 million images daily.
IT MAY INTEREST YOU: They will notify seismic alerts on cell phones starting in 2023 in CDMX
Since inputs are open to human consideration, many of the requests were related to offensive requests. For this reason, OpenAI responded that “Learning from real-world use has allowed us to improve our security systems, making wider availability possible today. In recent months, we’ve strengthened our filters to reject attempts to generate sexual, violent, and other content that violates our content policy, and we’ve created new detection and response techniques to stop misuse.”
MIDJOURNEY
The scale goes up a bit on impact when we talk about Midjourney. And it is that the level of detail and realism that it presents is remarkably higher than the rest of the options.
Its operation is similar. You must introduce specific concepts, as well as a series of visual parameters for the image to be generated.
One of the most surprising things is that some people used this tool to try to recreate visual styles of well-known artists. Whether in photography, film or video games, the internet is full of examples like the one you will see below:
As we already mentioned, many users started generating movie poster frames, and that’s when they discovered a tool that specifically does this: This movie doesn’t exist.
This allows you to generate random posters, or a more personalized one by entering a name, synopsis and a genre.
STABLE DIFFUSION & RIFFUSION
After all the cases that were seen in the first half of the year, in August 2022 the new app that caused a sensation was launched. And although it may not have been as popular in the mainstream, it was among digital artists. This is Stable Diffusion, which stands out, in addition to the versatility in the images created, for the integrations and plugins that allow you to expand your own limits.
What is the secret? The open source.
The following is an image that was generated under the input: “a mecha invades Mexico and the guy from 8 comes out in his defense.”
The big surprise came at the end of the year when Seth Forsgren and Hayk Martiros developed rifusion, a platform that can generate spectrogram images that can in turn be converted into audio. In other words, it converts images into sound.
On the official page you can find some examples and an extremely detailed explanation of how the software works, as well as the code itself.
LENSA
The last of the applications that attracted attention in 2022 was Lensa. If you did not get to see images generated by the previous examples, it is very likely that you have seen them at least once on social networks.
The difference with the other tools is that this one is not based on abstract external concepts, but is based on a selfie.
Lensa describes itself as a “photo editor that allows you to touch up photos with a myriad of effects”.
IT MAY INTEREST YOU: ‘Chocolate cars’ affect new vehicle sales, warn dealers
The criticism came when information spread that this application was stealing or using users’ personal data without authorization.
For its part, Prisa Lab, the developer of Lensa, declared that although the user retains the copyright on the content they generate: “You grant us a perpetual, revocable, non-exclusive, royalty-free, worldwide, fully paid up, transferable, and sublicensable license to use, reproduce, modify, adapt, translate, create derivative works from, and transfer your User Content, without further compensation. to you and always subject to your further explicit consent to such use where required by applicable law and as stated in our Privacy Policy)”.
It is not the first time that the fine print of this type of software has been questioned, as something similar had already happened with FaceApp, a tool that gained popularity by making supposed gender changes in images in 2020.
ANOTHER TYPE OF IMPLICATIONS
Everything that happened this year increased the debate on copyright, the necessary regulations for this type of tools, if the people who generate this type of images are artists or not, if the results produced by the algorithms can be considered racist or discriminatory. .
The truth is that these issues are still too much in the present to address them clearly, but what is certain is that the debate will go on and on and on. What do you think?
