DAY/TIME/PLACE
Asynchronous Online

Instructor: Bruce Donald Campbell

Faculty, Continuing Education - RISD
Providence, RI

Visual Analyst, CENIC
Providence, RI

Email: bcampbel01@risd.edu

Prerequisites: None

DESCRIPTION

The arrival of Artificial Intelligence (AI) technology has captured the attention of the creative industry with its ability to translate a prompt into compelling images—all generated from seemingly thin air! What does this powerful tool mean for the future of art? Get an introduction to how AI programs function and learn how to use this tool in your own creative process. You’ll also examine the cultural, creative and legal implications of AI imaging and what lies ahead.

GRADING

Class participation - 20%
Homework assignments - 40%
Written project - 40%

RESOURCES

Books

There is no specific book associated with this course. Readings are provided as PDF documents and supplemented by website references.

Course Handouts and On-line Readings as identified below and in class.

---

LINKS

• DALL-E 3
• Stable Diffusion Online
• ChatGPT
• Adobe Firefly
• Aitubo (based on Midjourney)
• MuseAI (based on Midjourney)
• Microsoft Co-Pilot
• NightCafe

SYLLABUS

 
Goal: To consider how artificial intelligence might inject into an artistic practice using mediated action theory as a lens.

By the end of this week you will: 

• have gained some exposure to the history of artificial intelligence methods (including machine learning and deep learning).
• understand how you might interject artificial intelligence into a successful artistic practice.
• have gained a basic understanding of how training by example, attention, transformation, and diffusion work to generate artificial content.
  

Consider that a key component to genAI strategy implementation is the hardware component called a Graphics Processing Unit (GPU). That technology was first researched and pursued from those interested in computer graphics. It has grown to support all kinds of computation where numerical data can be partitioned into pixel-equivalents.

Reading:

1. How Text-to-Image Generative AI Is Transforming Mediated Action (PDF)

Examples:

• To be provided

Significance to our course:

In the early 1990s, mediated action theory was offered up by sociologists as a way to think about why and how human beings perform actions. Not surprisingly, coming from sociologists, the theory suggested culture had a big impact on how humans behaved. An explosion of software use has occurred as humans turn to computing tools as a bigger percentage of the tools they use on a regular basis. We begin our course with a first topic that is intended as a level set introduction to artificial intelligence, but offer up an opportunity to use mediated action as a lens by which we think through the cultural ramifications and how they might influence our artistic practice.

Is artificial intelligence just another tool along the lines of other software we've used to help us realize our artistic vision? Or is artificial intelligence more an embedded aspect of the object we hone in on when pursuing an iterative approach to generating it? Let's expand on that even further and ask ourselves how artificial intelligence might be another subject we collaborate with as we might collaborate with another artist to co-produce a satisfying artistic result. Does mediated action help us expand our vision of how we might inject artificial intelligence into our artistic practice?

We will spend our first week of the course working together to explore genAI tools by way of iterating content with the use of them. As you will see in the Assignment page within this topic 1 module, we are going to daisy chain our iterations such that each class participant takes a turn in a round robin process whereby we see what we come up with as a class. As we do so, we'll keep the thoughts we get from the assigned reading on mediated action, provide above, in the forefront of our minds. Hopefully the discussion that ensues will suggest the individual contributions from each of us seem  to value more than the sum of contributions.

Perhaps we won't feel satisfied that we've iterated enough. If so, we'll continue iterating as long as we want, while pursuing our individual work in parallel. Hopefully that will provide a satisfying experience for every participant.

Homework:

Participate in the shared class exercise through iterative prompting, adjustment, and graphics tool manipulations as you think within the model of mediated action provided by the paper provided in the week's readings.

The instructions are provided in the demonstration video for topic 1. We are pursuing a daisy-chain ideation process whereby we assess the usefulness of social interplay with other people as part of our mediation with genAI tools.

The goal is for you to have an opinion regarding Midjourney's integration with required Discord channel(s).

Midjourney free trial

Use the topic 1 materials to journal your thoughts on mediated action, as they apply to your goals for exploring AI. Make notes regarding the different genAI tools you explored

_________________________________________________

If you have time left over, you can prepare for the technical discussion that will take place for topic 2:

AI methdologies (to be compared and contrasted in topic 2, but worthy of introduction now):

1. Symbolic AI, also known as Symbolic AI or Good Old-Fashioned AI (GOFAI), is a fascinating subfield that focuses on processing symbols and logical rules rather than numerical data
2. A* (pronounced “A-star”) is a powerful graph traversal and pathfinding algorithm widely used in various fields of computer science, and core to expert systems.
3. Neural Networks (NNs): Inspired by the human brain, NNs consist of interconnected nodes (neurons) that process information. Support tasks like image recognition and natural language processing.
4. Evolutionary Computation (EC): EC algorithms mimic natural evolution to optimize solutions. Genetic algorithms and genetic programming fall under this category.
5. Fuzzy Systems: Fuzzy logic deals with uncertainty and imprecision. It allows for gradual membership in categories (e.g., “very hot” instead of just “hot”)
6. Swarm Intelligence (SI): SI models collective behavior based on interactions among simple agents. Examples include ant colony optimization and particle swarm optimization.
7. Artificial Immune Systems (AIS): Inspired by the human immune system, AIS algorithms solve optimization problems by mimicking immune responses.

Useful AI-related concepts:

1. Gradient Descent:  An iterative method for finding a local minimum of a differentiable multivariate function.
2. Latent space is a lower-dimensional space that captures the essential features of the input data (in simpler terms, it serves as a compressed representation of the original data, where each dimension corresponds to a specific feature or characteristic).
3. Adjacency: Probability of things being in succession or proximity.
4. Diffusion Models and Transformers:
   • Probabilistic Technique: Diffusion models operate within a probabilistic framework, iteratively refining an initial image over multiple steps or time intervals.
   • Image Generation: They are specifically designed to generate lifelike, high-quality images.
     • Two Stages:
       •  Forward Diffusion: Random noise is gradually added to the training data, creating diverse samples.
       •  Reverse Diffusion: The models reconstruct data samples by undoing the added noise

 
Goal: To consider how artificial intelligence might inject into an artistic practice using data visualization practice as a lens.

By the end of this week you will: 

• have experience with using genAI tools with others in the context of mediated action theory.
• understand why style transfer is one of the successful milestones in the development of artificial intelligence.
• understand even more deeply how genAI methods depend on well-curated training data sets.

Reading:

1. Generative AI for Visualization: Opportunities and Challenges (PDF)

Examples:

• To be provided

Significance to our course:

Data visualization is a craft that is one part science and one part art. The craft is more bounded than the craft of creating art in general and the purpose of creating visual outputs is more precise: to communicate information clearly and effectively through graphical means. It's not just about making data look attractive but also about making it easier to understand and interpret. Data visualization can help to reveal trends and patterns, simplify complex data, support decision making, enhance communication, identify data errors, and facilitate exploration.

The science aspects of data visualization tend to draw in scientifically trained minds to the craft. Some of those scientifically trained minds have been contemplating the use of artificial intelligence in the process by which data visualizations are produced. They suggest the use of artificial intelligence can help with creativity, assist with production, and even automate production.

This week we consider the visualization workflow as a design case study that might provide insight into our own artistic practice. We have the paper referenced above for a deep dive of the components in this figure here:

Homework:

Participate in the shared class exercises that explore popular genAI tools (via cloud-based software).

The instructions are provided in the demonstration video for topic 2.

Use the topic 2 submission and discussion forum to submit examples of your work.

 
Goal: To better envision how we would use artificial intelligence in conjunction with the software tools we already use when making art.

By the end of this week you will: 

• rethink the nature of artwork when including software tools that assist in the creation.
• be able to describe how computer vision relates to machine learning techniques.
• have considered how artificial intelligence chaining of multiple applications may unleash surprising results.

Reading:

1. Jon McCormack: Art Infused with [Artificial] Intelligence (PDF)

Examples:

• To be provided

Significance to our course:

As artists, we already use many tools to create our art. Sometimes we like to think of tools broadly: language, penci, chisel, paintbrush, etc. Other times we like to categorize them to help us pursue the most useful ones. Graphic artists tend to use software to help create visual artifacts. Photographers tend to use software to enhance or modify their photographs in ways they consider create art.

This week we think about how we might inject artificial intelligence into the software tool workflows we are familiar with, or if we have no such workflows work with GIMP software to explore case studies we can learn from.

The goal is to do this in a group setting whereby we all become familiar with each other's current workflows, so as to better discuss and consider the potential of AI as a participant in the flow. Perhaps the software we use already has AI injected into the sevices provided by the menu choices we make. How do we feel about AI being injected into versions of familiar software? Is this like genetic engineering of our foods? Will we get upset if there is no transparent and clear labeling as to what is changing what's familiar?

An example of using DALL-E 2 and GIMP to create a montage of fictional book characters using prompts from their descriptions in a novel's text:

An example of generated plant flower designs, both at peak and at death of flower lifecycle, from genetic algorithm artificial intelligence:

Homework:

Participate in the shared class exercises that explore popular genAI tools (via cloud-based software) and then modify them with GIMP or other graphics manipulation tools.

The instructions are provided in the demonstration video for topic 3.

Use the topic 3 submission and discussion forum to submit examples of your work.

 
Goal: To consider computability and complexity as descriptors of artificial intelligence approaches.

By the end of this week you will: 

• have gained an opinion regarding hybrid aesthetics and their current state of the art.
• considered self-attention mechanisms for directing artificial intelligence methods.
• considered curatorship, authorship, and markets for AI-infused art.

Reading:

1. Testing the Capability of AI Art Generation Tools for Urban Design (PDF)

Examples:

• To be provided

Significance to our course:

During our early days in this course, we did an exercise where we incrementally added more complexity to a prompt used in a text to image AI-infused process. What effect does complexity have on computability? How complex can a prompt be and have a visual result represent each piece of that complexity explicitly?

And why do we use complexity as a metric in the first place? Are we reductionists who aim to reduce complexity to the essence of something we are trying to communicate? Or do we love to add complexity just to see how it perturbs the objects we are observing and studying, so as to better get a full picture of a phenomenon we want to communicate about?

AI is being lauded as a promising technology for helping us deal with complexity and gain insights into the nature of complex phenomena. The behavior of us, and our society, is one area where many are keen to understand so as to anticipate and solve problems caused by it. The statistical foundation for many AI approaches are sound, but recent learning models and deep learning techniques have been developed with baffling struggles of determining provenance.

The bottom line is that AI can both create and evaluate complexity. Can we get a sense of that through our own investigations of complexity, from which we can perhaps form our own surprising insights?

Discussion images from the PDF reading for topic 4:

Homework:

Submit a series of images that represent iterations and refinements towards your goal of supporting the short story provided in topic 3 with one to three illustrations. Submit any iterations on the story text you made when exploring dialogue with chatGPT.

Use the topic 4 submission and discussion forum to submit examples of your work.

 
Goal: To understand all the reasons for why artificial intelligence is useful for ideation and experiment with that purpose.

By the end of this week you will: 

• have formed some opinions regarding societal issues, copyright, and ethics associated with artificial intelligence.
• gain exposure to the benefits of iterative dialog with both AI-based chatbots and other human beings.
• be able to express your personal workflow that supports your most promising imagery to date.

Reading:

1. Mood Boards as a Universal Tool for Investigating Emotional Experience (PDF)

Other Resources:

• Investigate Mood Boards
• Consider Runway for your final project packaging.

Significance to our course:

GenAI tools have been lauded for their ideation potential. How can we integrate AI-assisted creations with our existing visual creation process? What text and visual artifacts help us summon up our creative process to consider what we want to create?

The evidence for mood boards as a useful aid to creative inspiration, as well as anecdotal evidence for it from many proponents, is strong so we'll all give mood boards a try to help expand our creativity in support of our final project work.

Let's explore our personal preferences for mood board generation and share our thoughts on how AI supported the process, given the tools we have explored so far in class. What conclusions have we reached so far with regards to AI's support of our ideation? Is this an ethically sound way to proceed on our way to creating our own artistic outputs? And how could the process become less ethically sound as we expanded upon it?

Example mood board created by a teen to support an intended mood for the month of February:

Homework:

Use genAI tools, along with any other software, to create a mood board that shares a mood along the lines of the benefits suggested by the PDF reading for topic 5.

Use the topic 5 submission and discussion forum to submit examples of your work.

 
Goal: To learn as much as possible from other artist workflows that incorporate artificial intelligence.

By the end of this week you will: 

• be able to express a vision for where artificial intelligence might evolve artistic norms and diversity.
• have submitted a self-directed project to spur discussion among your classmates.
• have participated in a design and interaction critique with a project of your own making.

This week you can search and seek all the resources you want to investigate given your project idea.

Be sure to bookmark all the resources you found useful from our course materials, and those that your classmates' suggested for you to explore that you haven't had time to explore.

Your attention can focus on providing suggested resources to your classmates. What did you find that helped you flush out your final project? Be sure to share them among the notes you submit that accompany your final project.

Significance to our course:

Hopefully our explorations in this course have been aided significantly through the process of exploring AI together in a guided group process? Let's do our best to communicate with each other so we can best determine if that's something we want or need to pursue going forward.

Various AI-generated images of a prompt consisting of text beaver in moonlight glow dream
(clockwise from top-left DALL-E 1, DALL-E 2, DALL-E 3, Stable Diffusion, Adobe Firefly):

Homework:

Finish up your milestone project so you can participate in our class critique. Be sure to document your workflow with clear explanations for how each image in your submission was produced.

Use the final project submission and discussion forum to submit your project work.