AI for Artists: A New Column by Swaptik Chowdhury for RabbleRouse News

“Swaptik Chowdhury is an AI fairness, ethics and safety researcher and is studying how AI technologies can be utilized to enhance public participation in technology and policy development.”

Part I. The Introduction

Generative AI or Gen AI applications have surged in popularity over the past few years, particularly following the public launch of ChatGPT in 2022. Since then, these technologies have found their way into various fields. For instance, computer programmers now use ChatGPT to write or complete code, while artists and designers turn to Midjourney to create custom art and user interface concepts. Corporations like Target and Best Buy have even adopted AI-powered chatbots to enhance customer service. (In fact, I’ve used ChatGPT to proofread and refine this very piece!)

In the artistic industry, tools like ChatGPT are being employed to write video scripts. At the same time, image generators such as DALL-E, Stable Diffusion, Midjourney, and Getty’s Generative AI are used to produce stunning visuals. Platforms like Suno are being used to create music and video generation is being explored through tools like Runway and Sora. Google’s NotebookLM can create hyper-realistic AI podcasts with some basic details. Notably, Coca-Cola recently launched a series of Christmas advertisements entirely generated by AI.

However, the rise of Gen AI is not without its controversies and skepticism. Concerns about privacy, the phenomenon of “hallucinations” (where AI generates incorrect information), and the environmental impact of these technologies are prevalent. The artistic industry is particularly grappling with worries about the appeal of cheap and “good enough” art, which could diminish entry-level job opportunities that artists typically rely on to develop their skills. Many production houses, such as Promise (backed by The North Road Company and Andreessen Horowitz) and Blumhouse, eagerly seek to integrate generative AI into their workflows.

Helping Artists Develop a Thorough Understanding of This Technology

Given the potential scale of Gen AI integration and the disruption it may cause, it is essential for the public—especially artists—to develop a thorough understanding of this technology. This knowledge will empower them to advocate for their interests more effectively. However, Gen AI is evolving rapidly, with a dizzying array of new developments emerging daily. Keeping pace with these changes and understanding their implications can be challenging. This difficulty is further compounded by the fact that much of the available information about Gen AI is filled with technical jargon, making it less accessible to the public.

In this column, I will try to “demystify” the technology behind gen AI and its recent developments. I will highlight notable advancements—whether products or technologies of generative AI—that all creatives should be aware of, from writers to actors to visual artists. The goal is to “co-create” knowledge and understanding with the artists about how these technologies may affect their work. While the primary audience for this column is the artists, it is an informative resource for anyone interested in the topic. Each week, I will introduce the standard fundamentals of AI and then explore the implications (good or bad) of any new advancement for artists through real-life examples.

This Week’s Focus: Understanding AI, ML, and Generative AI

This week, we will explore the concepts of Artificial Intelligence (AI), Machine Learning (ML), and Gen AI and the differences between them. People often use AI and ML interchangeably, which may lead to wrong conclusions about their impact or capabilities.

When we hear about AI, we often envision sentient robots from sci-fi movies poised to take over the world. In its most basic form, Artificial Intelligence refers to a machine’s ability to mimic human intelligence and perform tasks in the real world. Today, AI primarily achieves this through mathematical and logical processes, but this may change as new developments occur.

Machine Learning

Machine Learning is a branch of AI that focuses on ‘teaching’ machines to perform specific tasks. We do this by feeding the machine the data related to that task and identifying patterns and relationships in the data for future use. For example, predicting fraud in banking is an excellent example of understanding machine learning in action. In this example, the machine (computer code) sifts through many transactions or other financial data to spot common patterns in them, which are then associated with fraud. This process is known as ‘training.’ Once the machine has learned these patterns, we move to the ‘testing’ stage, where we fine-tune its ability using another data set. When the machine encounters new data and needs to determine if it’s a fraud case, it looks for the patterns it learned earlier. If it finds a match, it tags the transaction as fraudulent; if not, it labels it legitimate. This is called the “Inference stage.”

Generative AI, on the other hand, is a type of AI that utilizes deep learning as its mathematical foundation (or model). Deep learning is an advanced form of machine learning that involves processing “ginormous” amounts of data—like the 570 gigabytes of online text data used to train ChatGPT—to create mathematical “representations” that closely resemble/mimic the original data. Deep learning then uses the “representation” for prediction or classification tasks. But GenAI takes it further by using the “representation” used to create new data or content (More on this in the next column)
This brings us to an important issue: many authors, artists, and other creatives are suing generative AI companies for copyright infringement because these companies have “scraped” or collected vast amounts of text and images without permission, credit, or financial compensation to the original creators. The situation becomes even more complex when unattributed work is used to develop technologies that may replace authentic art with cheap and monotonous copies, ultimately undermining the value of genuine artistic expression.

Exploring Generative AI…

In the next column, we will explore generative AI and its different parts. We will also explore some AI tools that may help artists improve their workflow and finish mundane tasks.

I want to end this inaugural piece with a great quote from one of my favorite German philosophers, Theodor Adorno, who said, “Art respects the masses by standing up to them for what they could be, rather than conforming to them in their degraded state”.

Reach Out to Swaptik…

As this is supposed to be a bidirectional “co-production” of knowledge, I want to hear your thoughts. Please reach out to me at swaptikchowdhury16@gmail.com with your comments, questions or suggestions for topics to discuss.