AI asked to generate images of 'the last selfies ever taken' produces nightmarish results
People taking pictures of themselves with melting skin, blood-stained faces and mutated bodies, as they face a world on fire, will be the last selfies taken at the end of time, according to artificial intelligence.
The images were shared on TikTok by Robot Overlords who said they were simply asking to “show the last selfie ever.” However, it’s not clear which text-to-image AI system was used, as some commentators believe the eerie images were created with Midjourney.
The nightmarish results each show a human holding a phone and behind them are scenes of bombs falling, colossal tornadoes and cities on fire, along with zombies in the midst of destruction.
One of the selfies is an animated image of a man wearing something that resembles riot gear. He slowly moves his head around with a look as if his life is flashing before his eyes as bombs fall from the sky around him.
Each of the videos has been viewed hundreds of thousands of times, with users commenting on how horrifying each selfie is – one user felt the images would keep him awake at night because they are so hair-raising.
Scroll down for videos
A TikTok user asked a text-to-image AI to generate images of what he thinks will be the last selfies ever. One of the images shows a man in riot gear watching in horror as bombs fall behind him
Other users joked about taking a selfie at the end of time, saying, “But first, let me take a selfie” (if no one gets this reference, I’m going to cry).
TikTok user Nessa shared, “and my boss would still ask me to come to work.”
However, not everyone was light-hearted about what the end times would look like.
User Victeur shared: “Imagine hiding in the dark before the war, haven’t seen your face in years and see this when you take one last picture of yourself.”
The images were shared on TikTok by Robot Overlords, who said they were simply asking to show “the last selfie ever taken.” This result shows a horrified man who could run away from the devastation behind him
The latest selfies also show people with smeared blood and burning cities
Most commentators see the fun side of the images, but a dark side has been discovered with AI and specifically DALL-E – the racial and gender biases.
The system is public and when OpenAI launched the second version of the AI, it encouraged people to enter descriptions so that the AI can improve image generation over time, NBC News reports.
However, people started to notice that the images were biased. For example, if a user typed in CEO, DALL-E would only produce images of white men, and for “hostess” only images of women.
OpenAI announced last week that it was launching new mitigation techniques to help DALL-E create more diverse images and claims the update will make users 12 times more likely to see images featuring more diverse people.
The nightmarish images, showing zombies in front of burning cities, were created by a text-to-image AI system
Some of the disturbing selfies also look like a zombie with missing eyes and skin
The images are so chilling that some TikTok users said they will now have nightmares after seeing them
The original version of DALL-E, named after Spanish surrealist artist Salvador Dali, and Pixar robot WALL-E, was released in January 2021 as a limited test of ways AI can be used to represent concepts – from bland descriptions to fantasies.
Some of the early artworks created by the AI include a mannequin in a flannel shirt, an illustration of a radish walking a dog, and a baby penguin emoji.
Examples of phrases used in the second release – to produce realistic images – are ‘an astronaut riding a horse in a photo-realistic style’.
On the DALL-E 2 website this can be modified to produce images ‘on the fly’ including replacing astronaut with teddy bear, horse with basketballs and showing as a pencil drawing or as a ‘pop art’ in Andy Warhol- style painting.
DALL·E 2 has learned the relationship between images and the text used to describe them,” explains OpenAI.
“It uses a process called ‘diffusion’, which starts with a pattern of random dots and gradually changes that pattern into an image as specific aspects of that image are recognized.”
HOW ARTIFICIAL INTELLIGENTS LEARN WITH NEURAL NETWORKS
AI systems rely on artificial neural networks (ANNs), which attempt to simulate the way the brain works for learning.
ANNs can be trained to recognize patterns in information – including speech, text data or visual images – and are the basis for many advances in AI in recent years.
Conventional AI uses input to “learn” an algorithm about a particular topic by giving it massive amounts of information.
AI systems rely on artificial neural networks (ANNs), which attempt to simulate the way the brain works for learning. ANNs can be trained to recognize patterns in information – including speech, text data or visual images
Practical applications include Google’s translation services, Facebook’s facial recognition software, and Snapchat’s image-changing live filters.
Entering this data can be very time consuming and limited to one type of knowledge.
A new breed of ANNs called Adversarial Neural Networks pit the minds of two AI bots against each other, allowing them to learn from each other.
This approach is designed to accelerate the learning process and refine the output of AI systems.
.