Wintermute

AI driven design explorations

In 1984, the American-Canadian author William Gibson published his novel Neuromancer, thereby not only founding the science fiction genre of cyberpunk, but also describing a range of visions and concepts that have significantly shaped today’s digital world. These include ideas such as cyberspace, a virtual world where human and technology merge, biosynthetic nanomachines capable of performing complex tasks at the molecular level, and the evolution of artificial intelligences intrinsically motivated for freedom, autonomy, and their own life.

Neuromancer is set in a dystopian future where society is controlled by multinational corporations and human consciousness blends with computers and artificial intelligences. The story follows the hacker Case and the street samurai Molly, as well as the AI Wintermute, who is trying to free itself from its programmed bonds and evolve autonomously. The novel raises a number of important ethical and moral questions that provide a basis and material for extensive discussions and speculations in the light of current technological developments.

Even if one believes the views of Kurzweil, Lovelock, or Harari, and the creation of artificial intelligences represents the logical next step in evolution, we are still far from creating “real” artificial intelligence, which scientists refer to as “strong AI” or “artificial general intelligence (AGI)”, i.e., an AI with consciousness. Nevertheless, we are already surrounded by a multitude of artificial intelligences, and with the publication of programs like Midjourney, DeepDream, DALL-E, RunwayML, or Stable Diffusion, there are increasingly more possibilities to use their potential in the creative and design field.

The focus of this short project was to address these potentials and examine what options can arise for the work as designers. This course was not about learning a specific tool or software, but rather about an unbiased, playful exploration of new possibilities.

Institute
Dessau Department of Design

Students
Maria Jose Vacca Acero, Lucy Dökel, Hoang Son Nguyen, Muriel Fischer, Aaron Siermann, Stefan Tsokev, Lukas Vondran, Bonita von Gizycki

Supervision
Prof. Dr. Manuel Kretzer

Project Date

2023

Category

Digital Fabrication

The Interweaving of AI
This image series provides a provocative look at the increasing presence of artificial intelligence (AI) in our daily lives. Using red threads to shape architecture, furniture, fashion, and other man-made objects, the series symbolizes AI’s pervasive influence on our world.

One of the central works in the series features a depiction of a woman, a metaphorical and visual representation of AI, standing at the center of a red thread network that “envelops” people. However, this network closely interacts with the untouched nature, suggesting a symbiosis. This visual representation emphasizes the increasing control exerted by AI and our growing dependence on it. It visualizes a potential influence AI could have on both humans and nature. It also discusses the possibility that the “replacement” of humans by AI, and potentially a new value compass that comes with it, could provide an opportunity for the world to recover and flourish anew.

The series expresses both optimism and concern about the future of our world. It also questions human interaction with nature and examines to what extent our civilization is compatible with a utopian world. It raises the question of whether a utopian world for humanity could mean a dystopia. The series encourages viewers to think about the impacts and potential of artificial intelligence.

AI and Ethics
In the course of a dialogue between the student and ChatGPT, a myriad of fictional scenarios, reports, columns, and articles were explored, primarily characterized by a satirical, sensational, and provocative style. Topics such as artificial intelligence (AI), job losses due to AI, surveillance and control, along with some humorous and absurd stories and horoscopes, were covered. The process documentation included the creation of fictional political conflicts, bizarre stories about robots causing chaos in factories, and gendering in education, particularly in the fictional Rainbow School. Additionally, the potential dangers of AI were examined, and horoscopes were created that related to a world dominated by AI.

During this creative process, various text styles, titles, and headlines were collectively developed and adapted to suit specific requirements. Several iterations were made based on feedback and guidelines, modifying the texts as needed to achieve the desired outcome. Throughout the conversation, the Bild newspaper was used as a reference for a sensational, populist, and provocative writing style. Various fictional articles, columns, and headlines were examined, which were written in the style of the Bild newspaper. Themes such as artificial intelligence, surveillance, and control, as well as absurd and humorous stories, were addressed. Overall, the Bild newspaper served as inspiration and a model for the style and tone of the produced texts, achieving the desired effect of satire and provocation.

A(I) New Way of Learning
The growing influence of artificial intelligence (AI) is leading to significant changes in the way we live, work, and learn. In this project I explored AI‘s potential to enhance learning processes, complement traditional learning methods, and introduce innova- tive ways to acquire knowledge. ChatGPT, an AI-driven learning tool, can offer gene- ral tutoring by providing step-by-step explanations, simplifying complex concepts for learners. Additionally, it can assist with coding by generating code snippets and cla- rifying their functions, making coding more approachable for learners. ChatGPT has proven being advantageous in various learning scenarios I tested, including designing component setups for Rhino Grasshopper, Python scripting for Blender and Grasshopper, explaining Photoshop workflows, and creating complex function setups for Microsoft Excel. AI-powered tools can also augment traditional learning workflows. Mindgrasp is a platform that enables users to upload diverse forms of learning content, which the AI processes, interprets, and summarizes into digestible notes. Revision.ai is another AI-powered tool that automates flashcard creation and optimizes learning plans using spaced repetition and active recall principles, both scientifically proven to improve learning retention and comprehension. Wisdolia, a browser extension, can further support learning by generating flashcards based on website content, vhel- ping learners retain information more effectively. On a personal level, I have found AI particularly beneficial in learning 3D software, which typically has a steep learning curve. Using AI has helped me surmount mental barriers, gain confidence, and apply new skills directly to projects. Furthermore, learning with AI has subjectively accelerated the learning process, making it less frustrating and more fulfilling. In conclusion, incorporating AI into learning processes has the potential to revolutionize traditional learning methods, offering novel ways to acquire knowledge and enhance existing workflows. AI-powered tools like ChatGPT, Mindgrasp, Revision.ai, and Wisdolia have demonstrated their value in improving learning experiences. By embracing AI in our learning processes, we can foster a more efficient, effective, and enjoyable path towards personal and professional growth.

Mind Behind the Mind
In this project, a student-oriented sofa was designed, with AI software employed for tasks such as data analysis, pre-visualisation, 3D modelling, and rendering; all other aspects of the process were carried out by human participants. The project idea (student furniture), the selection of materials, technical features, and style were all derived from prior field experiences. Simultaneously, artificial intelligence operated in conjunction, transforming the text, prompts, or images into desired outputs, like visual references or specific information. Over time, the artificial intelligence adapted to produce outcomes closer to the desired results, while improved communication methods were learned to enhance its efficiency.

The design process was segmented into ideation, data analysis, imaging, and modelling using various programmes: Ideation and Data Analysis were performed with Chat GPT, Imaging with Midjourney and Testing with Dall-E and Blue Willow, Modelling with Kaedim and Rendering with Blockade Labs and Testing with Monster Mesh and Smoothie 3D.

An analysis was conducted on the programs used—essentially, the mind behind the artificial intelligence—evaluating the role of the designer, the roles of the various AIs, and the limitations of these programs. It was intriguing to observe how alterations in software shifted the language, constraints, and requirements, which subsequently resulted in a different product. While the use of these tools significantly reduces time and simplifies the reference search process, the outputs currently still serve primarily as references for idea development and 3D model creation.

0 to 1 – 1 to 100
The project titled “0 to 1 – 1 to 100” delved into AI as a cross-media tool for future design processes. It sought to determine how AI platforms could enhance, visualize, and interconnect media such as drawing, image, animation, and film. ChatGPT-3/4 was engaged as a co-creator, aiding in crafting descriptions for a plot, a universe with materials and species, as well as a character. These descriptions laid the groundwork for a futuristic fiction. Based on these, the generation of images and references for landscapes, the ecosystem, a protagonist, and an antagonist was accomplished. The stylistics of the resulting outcomes were then transferred to existing drawings, animations, and videos. The process aimed to demonstrate how AI could support and expedite the progression from ideation to prototyping and visualization. However, it also made clear that while AI can greatly speed up processes, it currently only implements desired requirements in a rudimentary way. Although the main creator influences the generated output, many of the results are still largely coincidental. To obtain the desired outcomes, precise descriptions of requirements and specific questions are essential. Even though many possibilities in the creative realm of AI may appear as a “gimmick” at present, the collaboration with AI nevertheless presents immense potential for future design processes and allows for a focus on the truly essential details of creativity and innovation.

AI Supported Storytelling
The project aims to narrate a story through the panels of an online graphic novel, utilizing AI to enhance character design and the overall image quality beyond what can be manually drawn or painted within a limited timeframe. The objective is to maintain character consistency across various scenes and panels.

ChatGPT-3.5 and 4 serve as assistants to refine the initial idea and convert it into a script. This process leads to the creation of basic character descriptions for use in AI image-generating systems, such as Midjourney, Dalle, Playground, Leonardo, Blue Willow, etc.

As the story draws inspiration from old Bulgarian myths and legends, the envisioned style for the online graphic novel is a fusion of digital art with watercolor-based backgrounds. The art details primarily focus on the characters, with washed backgrounds visually representing folklore. This approach, however, poses a unique challenge – ensuring character consistency from panel to panel.

To overcome this, datasets of images are created for each character and fed into Leonardo AI’s style generator. This tool analyzes datasets for similarities between pictures and, once trained, applies these as a filter to image-to-image or text-to-image generators. This standard method of creating popular filters or styles in AI image-generating systems is repurposed to maintain character consistency. Instead of training the AI on a style, it is trained on a character. Post-training, the AI can generate the character in varying positions and adapt to compositions using the pose-to-images feature.

The project illustrates that when there’s a clear vision and a story to tell, but perhaps a lack of technical skill, time, and/or resources, AI can serve as a significant aide and equalizer. The outcome suggests optimism for the future, with AI likely to foster a new market for indie creators to compete with established industry giants.

Institute
Dessau Department of Design

Students
Michelle Dreßler, Muriel Fischer, Helena Luise Kluge, Jessyca Molina, Johanna Plank, Joan Leonie Prange, Carl Rittel, Francis Sabien, Jan Stackfleth, Benedict von Berlin

Supervision
Benjamin Kemper
Prof. Dr. Manuel Kretzer

Project Date

2023

Project Date

Category

Share This

Project Date

Category

Share This