Contents:

How to make money quickly in IT in 2025 Try your hand at different areas of IT and take the first step toward a new remote career!
Learn moreArtificial intelligence is capable of many things: it can hold conversations, provide medical care, cook meals, and even create pictures of cats. However, to truly scare or surprise people, it takes more than just technology; it requires a creative approach and a deep understanding of psychology. We learned this lesson as children, especially those of us who spent time at summer camp.
Scientists have made significant advances in creating neural generators that generate amazing and terrifying images. We invite you on a thrilling journey through the darkest and scariest aspects of data science. Expect moments that may seem creepy. Be mindful of your emotional state.
In a pitch-black house there is a pitch-black room…
A shadow behind me. I felt a hand grab my ankle and pull me deeper into the dust. This could be the beginning of a new gripping story by Stephen King. However, this text is the product of a neural network. In a world where artificial intelligence increasingly uses literary devices, the boundaries between human and machine authorship are blurring. Neural networks are capable of creating captivating stories, provoking discussions about the future of literature and creativity.
Scientists from the MIT Media Lab, who specialize in artificial intelligence, celebrated Halloween in an original way in 2017. In honor of this holiday, they presented the Shelley neural network, designed to create scary stories. The development was named after Mary Shelley, the famous author of the Gothic novel Frankenstein. The Shelley neural network demonstrates the potential of AI in the field of creativity and literature, combining technology and art to create unique content that may interest fans of the horror genre.

To implement the project, programmers compiled a dataset of 140,000 creepy stories published on Reddit. In particular, in one of the discussion threads called NoSleep, users share their own horror stories, reminiscent of those we told friends on short June nights. These stories have become a source of inspiration for many creative projects and research in the fields of literature and the psychology of fear.
Dozens of cult horror novels were added to these stories, which became the basis for training the Shelley neural network. Using deep machine learning algorithms, this neural network has learned to analyze the structure of horror works and reproduce characteristic elements when creating new texts. This allows Shelley to generate unique stories, preserving the atmosphere and tension inherent in classic horror works.
Shelley, an artificial intelligence, is not yet capable of independently writing a novel or even a short story from start to finish. In response, it was posted on the @shelley_ai Twitter account, where anyone can help AI create new horror films. This initiative brings together human creativity and technology, opening up new horizons for the horror genre. Users can suggest ideas and storylines, making the creation process unique and interactive.
The collaborative creativity between people and the neural network was accomplished as follows. First, the Shelley AI generated a short text fragment of up to 140 characters, tagging it with the hashtag #yourturn. Users then retweeted this fragment and suggested their own continuations. The neural network selected the most suitable option and, based on it, generated the next text fragment. The result was a unique bouts-rimé format inspired by the image of Freddy Krueger. This process demonstrates how human and AI collaboration can lead to creative and interesting results.
Shelley worked for about a month, publishing new text fragments every hour. Pinar Yanardag of the MIT Media Lab noted that the key to the AI's performance was the use of powerful graphics cards for both training the neural network and its subsequent operation. These graphics cards provide high processing speeds, allowing the AI to generate text materials with greater efficiency and quality.
Shelley must quickly respond to Twitter user suggestions, making speed a key factor. This efficiency is ensured by the use of NVIDIA Titan X graphics cards, which help process data with high performance.
The collaboration between humans and AI resulted in the creation of more than 450 gripping and tense stories, requiring only minimal literary revision. Experts noted the high quality of the resulting works, many of which are comparable to the work of contemporary horror authors. These stories open new horizons for writers and fans of the genre, demonstrating the possibilities of synergy between human creativity and technology.

The story, compiled on the basis of texts generated by the Shelley neural network, is a unique experience in the field of artificial intelligence and literature. The Shelley neural network, designed for creating literary works, is capable of generating original texts that explore a variety of topics and genres. These texts demonstrate both creativity and technological advances in natural language processing. Examining the works created by Shelley opens new horizons for understanding the interaction between humans and machines in the field of creativity. Each generated text is the result of complex algorithms and the analysis of large amounts of data, making them interesting for both readers and researchers. The use of neural networks in literature continues to expand our understanding of creativity in the modern world and offers new perspectives for future research in this field.
In the silence of my bedroom at night, I felt a growing anxiety. Fear gripped me, and I didn't know how to cope. I realized that someone was watching me, and in the darkness, I noticed vague outlines. A chill ran down my spine, intensifying the feeling of anxiety and uncertainty.
I felt something reaching out to me. There was no one around, and the creature barely resembled a human. Its hair was pulled back into a bun and wrapped in a bloody towel. Its face was contorted in a grim expression, reflecting fear and suffering. At that moment, I realized I was in a strange and frightening situation, where every detail could matter.
She pulled out a photograph with illegible scribbles on it. I snatched it from her hands. It was a picture of me and an unfamiliar woman. I couldn't recognize her.
I froze in fear, unable to move or speak, even though the urge to scream for help was raging inside me. Panic gripped me, but my body wouldn't budge. The blanket I was under was soaked in blood, and it was slowly trickling down my legs. I felt a cold breath on my cheek, and my whole body shook with terror.
I pinched myself and felt something touch my shoulders. Suddenly, darkness enveloped my eyes. When I came to, I found myself in my bed. Lying there, I tried to recall the dream I had before losing consciousness. This moment marked the beginning of my reflections on what happens in the world of dreams and how they affect our perception of reality. Dreams can be mysterious and sometimes frightening, but they always leave a trace in our consciousness.
Of course, I am ready to help. Please provide the source text, which needs to be edited and optimized for SEO.
During the implementation of the project, both hardware and software solutions from NVIDIA were likely used. In particular, the NVIDIA CUDA (cuDNN) library of deep neural networks was used, which significantly optimized computational processes and increased model performance. The use of cuDNN ensures the efficient execution of operations required for training and inference of neural networks, which is a key aspect of the successful implementation of the project.

Shelley is a unique combination of a multi-layer recurrent neural network and an online learning algorithm. This approach allows her to adapt and improve her skills based on user feedback. The more Shelley engages with her audience, the better and more compelling her stories become. Through continuous learning and interaction with people, Shelley is able to generate original and engaging stories, making her an outstanding tool for lovers of literature and creativity.
Pinar Yanardag is a principal researcher at the MIT Media Lab. His work focuses on the interaction of technology and society, analyzing how new media formats influence user behavior and perceptions. Yanardag is actively involved in developing innovative solutions aimed at improving quality of life through technology. His research covers a wide range of topics, including digital culture, social media, and the ethical aspects of technology. MIT Media Lab, known for its cutting-edge ideas, provides a platform for such research and promotes the development of interdisciplinary approaches in scientific endeavors.
MIT Media Lab researchers believe that Shelley will become the basis for future artificial intelligence models capable of generating content that evokes specific emotions, such as fear. This development could have a significant impact not only on literature and journalism, but also on marketing and advertising, opening up new horizons for creating emotionally charged content. The use of AI in these areas will enable more effective audience engagement and create memorable experiences.
The AI was small, and Santa turned out weird
Inspired by the work of their colleagues in the field of data science, journalists from MIT Technology Review, Will Knight and Karen Hao, decided to experiment with neural networks. Ahead of 2019, they set themselves the task of training artificial intelligence to create original plots for Christmas films. This project not only demonstrates the capabilities of modern technology but also opens new horizons for creative content in the film industry.
To achieve this goal, they collected synopses of 360 New Year's blockbusters from Wikipedia and loaded them into the textgenrnn recurrent neural network. This neural network is capable of analyzing and generating textual information, enabling it to create new and original descriptions based on existing data on popular New Year's films.
After training, the artificial intelligence generated numerous stories, but most of them were incoherent jumbles of sentences. Only a few texts contained elements of common sense, but they resembled the plots of horror films more than lighthearted New Year's comedies. These stories frequently mentioned terrorists, murders, and fights, which clearly did not fit the holiday atmosphere.
The scientists whom Knight and Hao contacted explained the reasons for the shortcomings in their study. Firstly, a dataset of 360 records is too small for fully training neural networks. By comparison, Shelley used 140,000 examples, which significantly increases the quality of text generation. Secondly, neural networks such as textgenrnn are not able to adequately assess the plot consistency in the long texts they generate. Shelley's creators partially solved this problem by breaking stories into fragments and engaging Twitter users as co-authors. This collaboration helps improve the quality of the generated content and makes it more consistent and logical.

The neural network forms each subsequent word based on the probability of its appearance after the previous one. This process can be compared to using T9 technology when writing text. This runs the risk of losing logic and grammatical consistency. Neural networks, although capable of generating text, often face difficulties in maintaining a coherent and logical narrative, which can lead to unclear and confusing results.
Karen Hao, a journalist at MIT Technology Review, provides insightful analysis of the latest technological trends and their impact on society. Her articles highlight key aspects of innovation, exploring how technology is shaping the future. Hao focuses on ethical issues and social implications associated with the development of new technologies. Readers can expect reliable data and insightful commentary that help them understand complex contemporary issues.
Knight and Hao went further and decided to use artificial intelligence to solve a simpler task: generating titles for new movies. As a result of their efforts, the textgenrnn network was able to generate several dozen original titles, perfect for New Year's action and horror films. This discovery demonstrates the potential of AI in creative fields and its ability to inspire new ideas in the film industry.
The journalists identified three key films: "It: Santa," "The Christmas Mile," and "Fight Christmas." As part of the study, the authors asked designers who don't use artificial intelligence to create posters for these films. This highlights the importance of human creativity in the visual presentation of cinema and its role in promoting the New Year theme.



Horror Machine
Elon Musk, discussing potential threats from technology, noted that by developing artificial intelligence, humanity could face serious risks. In response to his warning, researchers at the MIT Media Lab decided to create a project known as the Nightmare Machine. This innovative tool aims to explore the boundaries of artificial intelligence and its impact on society, raising important questions about the safety and ethics of technology.
A group of scientists led by Professor Iyad Rahwan, as part of the Nightmare Machine project, developed an artificial intelligence system capable of transforming photographs of cities into terrifying post-apocalyptic landscapes. The developers claim that this is the world's first AI designed to evoke fear in people. The project opens new horizons in the study of emotion and perception, demonstrating how modern technologies can influence human feelings and perception of reality.
American researchers applied the Nightmare Machine algorithm to processing images of famous tourist attractions. These objects included the Eiffel Tower, the Statue of Liberty, the Taj Mahal, the Colosseum, and St. Basil's Cathedral in Moscow. The algorithm created eight unique "nightmare" patterns, each given a memorable name: "Haunted House," "Fright Night," "Slaughterhouse," "Toxic City," "Ghost Town," "Hell," "Tentacled Monster," and "Alien Invasion." These patterns represent distorted and frightening versions of famous locations, offering a unique and disturbing perspective on familiar tourist spots. The study highlights how technology can transform our perception of the world around us and compel us to explore the hidden aspects of familiar cultural icons.

Nightmare Machine is based on Neural Style Transfer (NST) technology. The first successful NST algorithm was presented in a 2015 scientific paper by researchers from the University of Tübingen in Germany. They demonstrated how artificial neural networks can transform ordinary photographs into works of art stylized as works of masters such as Van Gogh, Edvard Munch, and Pablo Picasso. This technology opens new horizons in the field of digital art, enabling the creation of unique visual images and inspiring artists to new experiments.
The creators of the Nightmare Machine project used a convolutional neural network that uses the Neural Style Transfer (NST) algorithm. Thanks to its multi-layer architecture, this network effectively distinguishes between the content of an image (what is depicted) and its style (how it is depicted) during the training process. This allows you to create unique visual effects by combining elements of different styles and thereby transforming original images into new works of art.

Unlike their European colleagues, MIT Media Lab researchers trained their neural networks not on works by famous artists, but on still frames from horror films, science fiction, and reports on environmental disasters. This approach allows them to create unique visual solutions based on the emotional component and social problems of our time. Using unconventional data sources expands the horizons of artificial intelligence and opens up new possibilities for creativity and analysis.

We use deep learning algorithms to train a neural network to recognize images of haunted houses, abandoned cities, and landscapes polluted with toxic emissions. The algorithm analyzes training images, extracting stylistic elements, such as a somber color palette, and transfers them to ordinary photographs. This creates unique visual effects and improves understanding of different types of environments.
Pinar Yanardag is a principal researcher at the MIT Media Lab. His research focuses on the interaction of technology and society, as well as the impact of digital technologies on everyday life. Yanardag actively works on projects that explore how innovations can change the way we perceive and interact with information. Pinar's contributions to the field of media and technology continue to attract the attention of scholars and professionals seeking to understand how today's technologies are shaping the future.
Iyad Rahwan's four-person team was able to develop Nightmare Machine in just two weeks by leveraging existing technologies. This achievement highlights the importance of leveraging existing technologies and solutions in the rapid creation of innovative projects.






The MIT Media Lab team, led by Rahwan, continued working with the Nightmare Machine algorithms in a new project called Deep Empathy. This project aims to study and develop technologies that facilitate emotional interaction and understanding between people through digital platforms. Using algorithms from previous research, the team aims to create tools that will help users more deeply understand the feelings and experiences of others, a crucial step in developing emotional intelligence in the digital age.
As part of this project, the researchers trained artificial intelligence using photographs of Syrian cities devastated by the civil war. As a result, the neural network acquired the ability to transfer images of war-torn cities to images of cities around the world. This project highlights the importance of using AI to analyze and visualize the aftermath of conflict, and to better understand the scale of destruction and its impact on urban environments.

The project aims to awaken public empathy for the victims of military conflict. MIT Media Lab specialists used the Deep Empathy neural network to process photographs of cities such as London, Paris, Amsterdam, Boston, and Moscow. These images enable residents of prosperous cities to grasp the scale of destruction and suffering associated with war. This innovative approach to visualization allows for a deeper understanding of the consequences of conflict and promotes greater empathy for those affected by violence and destruction.
Deep Empathy was developed with the participation of UNICEF. The project page invites viewers to make donations to support UN initiatives aimed at helping Syrian children. Donations will go to projects that provide access to education, medical care, and basic resources for children affected by the conflict in Syria. Your contribution can make a significant difference in the lives of these children and help them restore hope for the future.
To improve the neural network's effectiveness, a survey was launched on the project's website, allowing users to choose the one of two photographs that evokes the greatest emotional response. This approach allows us to collect data that will help improve image processing algorithms and enhance the performance of artificial intelligence.



How to code ghosts and zombies
MIT Media Lab students Ziv Epstein and Matt Groh, continuing the research of their senior colleagues, have developed a neural network capable of generating images of ghosts. At the moment, this technology is used exclusively for creating photographs.
The project, which started as a harmless initiative, was conceived by Epstein and Groh to create artificial intelligence that could remove individual people from photographs and restore the background. This function had the potential to be useful for those who would like to get rid of images of friends and acquaintances in old photographs with whom the relationship had become strained. The development of such AI opens new horizons in image editing, allowing users to easily modify visual memories to fit their current lives.
The program developed by the students proved less effective than they expected. Instead of clear images, only blurred silhouettes remained in place of the removed objects. This prompted the students to consider using neural networks for the inverse problem—not removing, but adding fuzzy, ghost-like figures to images. Thus, they opened new horizons for the use of image processing technologies, which could lead to interesting and creative solutions in digital art and design.
The "AI Spirits" project, presented by Epstein and Groh in 2018, marked a significant step in the field of artificial intelligence and generative art. Using generative neural networks, AI Spirits can process photographs of empty landscapes—whether a seashore, a dark forest, or a wheat field—and add humanoid silhouettes and shadows. This process creates unique visual effects, evoking a sense of presence and interaction with the space. The project illustrates the potential of AI to create new forms of art and expand the boundaries of perception of reality.



Having trained on a database of about 5,000 photographs of people, the artificial intelligence has mastered the basic rules of composition used by professional photographers. AI Spirits places ghostly figures in places where our brains expect to see them, such as on a forest path. Thanks to this, the phantoms created by the neural network are truly capable of causing fear and anxiety in viewers.

The model was trained by comparing photographs with images of people and similar images in which these people were removed using artificial intelligence. The algorithm integrates "ghosts" into the expected locations based on an understanding of the standard size and placement of human figures in images. The neural network colors and textures these ghosts based on the surrounding landscape visible in the rest of the photograph. This creates the illusion of presence, which could be useful in various fields, such as advertising, film, and digital art.
Matt Groh, a researcher at the MIT Media Lab, noted the importance of innovation in modern technology. His research focuses on human-computer interaction, which opens new horizons for the development of user experience. Groh emphasizes that innovative approaches in this field can significantly change our perception of technology and its impact on everyday life. A quote from an interview on the Digg platform highlights his desire to create more intuitive and accessible interfaces, which is especially relevant in the context of rapid technological progress.
The young scientists presented their invention to demonstrate the potential of artificial intelligence, which many still underestimate. Epstein and Groh note that neural networks, as they train, are capable of acquiring unique traits of human nature that even humans themselves may be unaware of. This discovery highlights the importance of AI in our society and its ability to reveal hidden aspects of human behavior.

Artificial intelligence is not a malevolent force, but its capabilities can be concerning, especially when fed the wrong data. Without proper controls and oversight, the use of machine learning models can lead to serious negative consequences. It is important to understand the risks associated with the implementation of AI and develop strategies to ensure its safe and ethical use.
Matt Groh, a researcher at the MIT Media Lab, actively studies emerging technologies and their impact on society. His work focuses on human-machine interaction and ways to improve user experiences. Groh emphasizes the importance of digital media innovation and its potential to transform our approach to information and communication. Research conducted at the MIT Media Lab aims to create more intuitive and accessible technologies, which can significantly improve quality of life.
Besides ghosts, neural networks are also capable of creating zombies. This is possible thanks to the Nightmare Machine algorithm, developed at the MIT Media Lab. This tool can apply a terrifying style not only to images of cities, but also to photographs of people, turning them into the walking dead. This demonstrates how modern technologies can alter the perception of reality and create unique visual images.
To train artificial intelligence in this complex skill, scientists used photographs of famous personalities. The neural network was trained by applying zombie features to the images. However, the researchers did not disclose the source from which they obtained the required number of photographs of the undead to train the AI. This study demonstrates how modern technologies can combine elements of art and fiction, opening new horizons in the field of computer vision and image processing.




MIT Media Lab programmers, exploring the capabilities of generative adversarial networks (GANs), have developed a unique model, described in a scientific paper. This network has learned to create abstract faces that, when overlaid with zombie-like style, resemble characters from cult films by director George Romero, as if they've escaped their gloomy graves for one last selfie. The study demonstrates the potential of GANs in the field of visual content generation and its application in art and entertainment.

The algorithm first learned what a face looks like, which was achieved by processing 100,000 images of celebrity faces. After that, the algorithm learned to generate new images on its own. When creating unique faces, Nightmare Machine uses a special algorithm that allows it to "zombify" these faces, imitating the characteristics of a real zombie. This combination of technology and creativity opens new horizons in the fields of image generation and animation.
Pinar Yanardag, a leading researcher at the MIT Media Lab, actively studies modern technologies and their impact on society. His work covers various aspects of media, digital communications, and innovative solutions, making him an important figure in the field of research. Yanardag focuses on how technology can improve quality of life and facilitate social change. As part of his research, he analyzes human interaction with digital platforms and their impact on everyday life. MIT Media Lab, known for its cutting-edge developments, provides a platform for the implementation of his ideas and projects.
The project uses two neural networks. The first network is responsible for generating realistic human faces, while the second network transforms these faces into monsters. Using such technologies, it is possible to create unique and diverse images for various purposes.
Researchers have launched a voting system on the Nightmare Machine website, allowing anyone to help train the "Nightmare Machine." Users select the most terrifying images from a selection of ten, after which they receive a "reward" of a selection of 36 images that, according to the neural network, have the greatest impact on their perception. This interactive participation helps improve the algorithms for generating shocking content and provides a deeper understanding of which visual images evoke strong emotions in viewers.
Voting results play a key role in training the neural network, contributing to the creation of increasingly realistic and impressive images. Each vote modifies the AI algorithm, directing it toward certain characteristics: increasing the number of teeth, lightening the skin, or darkening the background. Over two million people have already participated in the voting, highlighting the high level of interest and engagement in the development of this technology.
Do the AI Spirits and Nightmare Machine projects have practical applications? Their creators insist they do. They emphasize that these works are designed not only to evoke fear, but also for more serious purposes. The projects aim to explore the relationship between artificial intelligence and human emotions, as well as to explore the limits of perception. The creators believe that such experiments can deepen our understanding of technology and its impact on society.
The research focused on the reasons why people develop fear in response to certain images. Understanding these factors could play a significant role in reducing the uncanny valley effect that has plagued robot and AI developers for years. This problem has a significant impact on the perception of technology and its interaction with people, which makes it relevant for further research and practical application.

The uncanny valley phenomenon has been studied by scientists for a long time. He describes the feelings of fear and disgust that people experience when interacting with robots that look almost like real people, but are still noticeably different. We therefore wondered whether artificial intelligence could deliberately evoke similar negative emotions, similar to those experienced by viewers while watching horror films. The results of our study provided a clear answer to this question.
Pinar Yanardag is a principal researcher at the MIT Media Lab. His research focuses on modern technologies and their impact on society. Yanardag actively works on projects that explore the interaction of people and technology and their role in shaping cultural and societal change. At the MIT Media Lab, he makes significant contributions to the development of innovative solutions that can improve quality of life and empower users.
Please note that you may have difficulty falling asleep today. We have warned you in advance.
Read also:
- What are you?! Neural Network Features That Surprised Even Their Creators
- Killing the Matrix in the Bud: 7 Arguments for Arguing with an AI Fan
- What is Machine Learning and Is It Worth Studying?

