Designer and AI: a new balance | Alexey Lysogorov | TEDxUCANaryn

Introduction and Background

• The speaker is a graphic designer with awards in the field of graphic design, and has worked on logos, brands, and corporate styles that can be seen in stores and markets in Kyrgyzstan (22s).
• Some of the notable works include designs for companies, the logo for the American University of Central Asia (AUCA), and navigation designs for the AUCA campus (1m3s).
• The speaker also taught at AUCA and has a connection to the building, which is reflected in their design work (1m11s).
• The speaker wants to show the audience something interesting, specifically Soviet-era mosaics in Bishkek, which include graphic design and stylization elements (1m38s).

Generated Mosaics and Deception

• The speaker asks the audience if anyone has seen any of these mosaics in Bishkek, and reveals that they actually generated the images using neural networks (2m6s).
• The speaker notes that it is easy to be deceived by generated images, and that they were able to trick the audience into thinking the mosaics were real (2m36s).
• New visualization tools are a topic of concern for many students who fear for their future in the profession (2m40s).
• A mini-research was conducted by walking around the city, photographing buildings, and creating real mosaics, but the presented images were actually generated using knowledge of styles and directions (2m55s).
• The images were created by taking real publications from Bishkek and making a montage, which is not a difficult task and can be done with basic photo editing skills (3m11s).
• Many people are deceived by the idea that these generated images are real, and this is a common issue with the use of neural networks and other visualization tools (3m24s).

Introduction to Neural Networks and Visualization Tools

• Popular neural networks used by designers and others are tools for visualization, and some examples include Deep Dream Generator and Prisma, which can be used for free with a Google account (3m37s).
• These tools can generate a limited number of images per day, but they can be used to try out different ideas and see what can be created (3m53s).

Example of Generated Image and its Imperfections

• An example of a generated image is a picture of a beautiful autumn scene in Kyrgyzstan, which was created using a simple prompt and a mix of balanced and classy images (4m4s).
• However, the generated image is not perfect, and there are some issues with it, such as a strange-looking cow in the scene (4m35s).
• Nature is not perceived as strange, despite its unusual generation, which alters proportions and balance, because it does not have a clear structure or a correct anatomy, and this is what makes it beautiful (5m0s).
• In contrast, living beings have a known anatomy, and any imbalance in their structure is noticeable, such as the number of legs a cow should have and their position (5m11s).

Early Experiments with Image Generation

• A series of experiments was conducted using old, approximately two-year-old works, when image generation was a novelty, and the results were very interesting, especially when generating realistic images of people (5m41s).
• An experiment was conducted using a fantasy concept of "Kyrgyz Androids," which involved combining traditional Kyrgyz ornaments with futuristic elements (5m50s).
• Another experiment involved using a prompt to generate an image in a Kyrgyz style, specifically "Arco" or "Nua," and the neural network was able to understand the style and direction, as well as recognize some cult authors and artists (6m32s).
• The neural network was also able to recognize the work of certain cult film directors, despite them not being painters, showing its ability to understand different artistic styles and directions (6m44s).

Balance in Work and AI Integration

• The concept of balance in work is crucial, and it's essential to find a balance when using AI, as it can be considered a part of one's own style, but not entirely one's own (6m53s).
• A balance in work allows for the creation of unique content, such as a fictional representation of Kyrgyzstan's "Zabroshki" (abandoned places) that looks realistic, making people believe it's real (6m58s).
• Using photo editing software like Photoshop, one can create visually appealing content that people are willing to believe, especially when shared on social media platforms like Instagram (7m26s).
• Creating entertaining content can be useful, as it allows for experimentation with new tools, learning, and self-improvement, and can also be a way to have fun and showcase one's creativity (8m3s).
• Experimenting with different styles, such as those of famous artists and sculptors like Antonio Gaudi, can help create unique content and add a personal touch to one's work (8m27s).
• Adding a personal touch to AI-generated content can make it more interesting and unique, allowing for the creation of something entirely new and original (8m40s).
• Finding a balance between AI-generated content and personal creativity is essential, as it enables the creation of content that is both visually appealing and unique (8m47s).

Pink Floyd Album Cover Project

• A project was created that gained a large international audience, particularly on Facebook, by imagining Pink Floyd releasing a new album dedicated to Kyrgyzstan (8m49s).
• The project combined different elements, including the work of Storm Thorgerson, a designer known for his work with Pink Floyd, and the style of René Magritte (9m13s).
• To create the project, knowledge of art history and design history was necessary, even when using neural networks, as it allows for a better understanding of the styles and inspirations of famous designers (9m31s).
• The project involved writing prompts for a neural network, including "Storm Thorgerson inspired by René Magritte," to generate album covers in a unique style (9m47s).
• The results were a mix of different styles, creating interesting and unexpected album covers (10m15s).
• The project was shared online, and some people believed the album covers were real, showing how realistic the generated images were (10m18s).
• The project demonstrates how AI can be used to create new and interesting content, and how it can inspire users to think creatively and become creators themselves (10m31s).

Using AI for Presentations and Design

• However, it's also important to consider the limitations and potential risks of using AI-generated content, as it can sometimes lead in unintended directions (10m49s).
• The process of selecting already generated works takes a long time, involving a large number of generations, and choosing the best option is a significant task (10m54s).
• The next step is to use the selected work for presentations, but designers often lack high-quality, unique images that fit their design and do not repeat or come from the internet (11m33s).
• To address this issue, neural networks can be used to generate images for presentations, not for the design itself, but for its presentation (11m50s).
• For example, the background images, or "maps," used for logos can be generated by a neural network, eliminating concerns about mixing other people's images and ensuring the design is unique (12m6s).
• This approach allows for a wide range of options, and the generated images can be ideal for the design, making it stand out (12m30s).

Neural Networks in Design Workflow

• The use of neural networks in design is not new, and designers have been utilizing this method since neural networks emerged, as seen in past projects (12m43s).
• One such project, created last year, relied heavily on neural networks due to a limited budget and the need for rapid execution, making it impossible to draw everything from scratch (13m3s).
• The process of working with neural networks involves selecting a style, and the network generates an image, but the output is not exactly what is desired, requiring further work and refinement (13m21s).
• The generated image serves as a reference, similar to how designers work with references, and this process can be repeated to achieve the desired result (13m36s).
• The use of neural networks can save time and resources, as generating an image takes only a minute, but the result may not be perfect and requires additional work, such as editing in Photoshop (14m24s).
• The goal is to find a balance between the work done by the neural network and the work done by the designer, with the understanding that the output will not be 100% ready and will require some manual work (13m56s).
• The process of working with neural networks is not a replacement for human work, but rather a tool that can aid in the design process, and it is essential to have a balance between the two (14m52s).
• The output from the neural network can be used as a reference, but it is not a finished product, and designers should not expect to receive a ready-made result, but rather a starting point for further work (14m47s).
• The use of neural networks can be beneficial, but it is crucial to understand that it is a tool that requires human input and refinement to achieve the desired outcome (15m4s).

Commercial Project: Vegetable Product Labels

• A commercial project for labels was created with the help of a neural network, as the individual is not a skilled artist, but knows the desired style and has a clear imagination of how the final product should look (15m18s).
• The project involved creating a catalog of characters for a brand of vegetable products, which would have required a lot of time and help if done manually, but the neural network was able to assist in the design process (15m48s).
• The neural network was helpful in generating ideas and designs, but sometimes struggled with creating complex products, such as a combination of a tomato and cucumber, resulting in unbalanced and awkward designs (16m11s).
• In such cases, it's sometimes better to hire a traditional artist, which is often done, especially when working with students who can create high-quality artwork (16m42s).
• However, with enough references and a clear idea of the desired style, it's possible to create a good design using a neural network, even with limited drawing skills (16m54s).

vit Milk Brand Logo Design

• Another project involved creating a new brand of alternative milk, which is not yet released, and the neural network was used to generate designs for the product (17m11s).
• A new milk brand called "vit Milk" is being developed in Kyrgyzstan, with flavors such as oat and coconut milk, and the brand's logo features the letter "V" made from splashes of milk with nuts inside (17m29s).
• The idea for the logo was conceived quickly, but its realization took about a week, as it required finding an executor, explaining the concept, and reviewing sketches (17m51s).
• The convenience of using neural networks allows for the creation of designs even through a smartphone app, making it a useful tool for generating ideas, such as during dinner or while watching a series (18m16s).
• A project for the "vit Milk" brand was developed, but it has been put on hold, and its release is still pending, although all the necessary elements are ready (18m40s).

Visualizing Ideas with Neural Networks

• Neural networks enable the rapid visualization of even the craziest ideas, such as imagining Master Yoda or Darth Vader taking a selfie, making it possible to instantly bring wild ideas to life (18m59s).
• The neural network can understand pop culture references, including Star Wars, and can generate images based on these references, as demonstrated by the creation of a Star Wars-themed image using the network (19m26s).
• The neural network can be used to generate ideas and designs even in leisure time, such as during a vacation, and can be a fun way to pass the time and explore creative ideas (19m18s).

Samsung Advertisement Project

• A project was created for Samsung, where the role was that of an art director and a balancer, with the task of finding a balance between the brand and the content (19m43s).
• The project involved creating a slogan "No matter who you are, no matter where you are, take a selfie" and a series of prints featuring characters from Star Wars, with the goal of creating a tolerant and interesting advertisement for the camera (19m48s).
• The project resulted in a series of 10 prints, but unfortunately, Samsung was unable to use them due to copyright issues, as the characters were too recognizable and the brand would have had to make agreements with Star Wars (20m11s).
• The project was more of an entertaining content rather than a commercial advertisement, and Samsung acknowledged it as "cool" and "genius" but ultimately unusable due to copyright concerns (20m23s).
• The main question raised by the project is whether it is tactful and correct to use such recognizable characters in an advertisement, and whether there is a balance between the brand and the content, with the conclusion that the balance was not achieved (21m13s).
• The project's outcome was a series of posters that were generated, showcasing the idea, but ultimately not usable for commercial purposes (21m32s).

Experimenting with the Blend Instrument

• The balance of composition in art is appreciated, and an experiment was conducted to generate art without a prompt, using an interesting new tool that mixes images, specifically the "Blend" instrument, which creates something new at the intersection of different images (21m57s).
• The Blend instrument was used to combine already balanced and great arts that are completely different, resulting in an experiment that is liked as a tool for creativity, allowing the user to decide what is good or bad and find a balance (22m17s).
• The works of Alexey's father, a designer and artist, were used as an example, specifically his designs for the first Soviet passport, and an experiment was conducted to combine two of his works using the Blend instrument (22m32s).
• The result of combining the two works was a third, new image that is interesting and provides food for thought, and it is unclear whether it is good or bad (23m13s).
• The experiment was shown to Alexey's father, who is a traditional artist but also uses modern tools in his work, and he evaluated the result as a "three", meaning it is not bad but not good either (23m43s).
• Alexey's father's reaction to the experiment was interesting, as he is a traditional artist who values his work but also appreciates modern tools and uses them in his creative process (23m36s).

AI in Art and Authorship

• An alternative evaluation of a person's work can be created using AI, as seen in the example of a blended image that is not the original work but rather a mix of different elements, making it difficult to distinguish what was created by a human and what was created by a neural network (23m59s).
• The neural network is capable of simulating handwritten signatures and titles, making it even more challenging to identify the creator of the work (24m42s).
• The ability to recognize what was created by a neural network is becoming possible, even in complex images, as people are becoming more familiar with the characteristics of AI-generated content (24m31s).
• An experiment was conducted by mixing two different artworks, resulting in a new image that combines elements of both, demonstrating the potential of AI as a tool for creating new and unique content (24m57s).
• The neural network is designed to balance the input elements to produce a visually appealing output, as seen in the resulting image that combines common elements such as a sun, face, knees, and a cat (25m28s).
• The use of AI in creating new content raises questions about authorship and copyright, as it is unclear whether the original authors of the mixed artworks would have any claims to the new image (25m56s).
• The creator of the new image may be able to claim inspiration from the original artworks, rather than infringing on their copyright, as the resulting image is a unique combination of different elements (26m7s).

The Evolving Concept of Style and Collaboration

• The concept of style is becoming increasingly collaborative, and when an image is slightly modified, the original author may not be able to claim it as their own, and may not even be able to find common ground with the modified version (26m14s).
• The modified image can be worked with as a reference, and the original author may not be able to say anything about it, and may not even be able to find anything in common with the modified version (26m23s).
• The question remains whether this development is good or bad, and opinions on the matter are still ongoing (26m42s).

The Future of Human-AI Collaboration

• The future of human-AI collaboration is uncertain, and it is unclear whether AI will simply be used as a reference or if it will eventually take over the work of humans (26m55s).
• The possibility of AI taking over human work is a potentially frightening prospect, but it can also be seen as a new and exciting tool that can be utilized (27m8s).
• The development of AI can be viewed optimistically as a new instrument that can be used, and it is not something to be feared (27m15s).