Home > News > The potential and pitfalls for AI in creative industries 

The potential and pitfalls for AI in creative industries 

Written by

Historians will remember 2023 as the year generative AI exploded into popular culture. It is hard to believe a piece of AI-created art stirred so much controversy by winning a digital art competition less than a year ago. These were simpler times. Now the controversies involving AI are far more serious, ever-increasing at a startling rate. AI’s rapid evolution has left news outlets struggling to keep pace with its changing uses and abuses across the web. AI’s impact is now undeniable, raising questions about how it will change society, work, and creativity. The key question we are asking here at MoMoLab is whether designers can find an ethical balance and use AI in positive ways for improving the quality of concepts and for enhancing creativity, while avoiding the minefield of negative externalities. 

Rise Of the machines

If we take a quick look at the past year’s rise of AI, we can shed light on the larger issue. When ChatGPT, a language model used to generate human-like text, first launched, students immediately recognized its potential as a tool for writing school assignments. Irony struck as ChatGPT was repurposed to spot AI-generated texts, triggering a cat-and-mouse game of deception as students attempted to continue cheating in ways undetectable by AI oversight tools. Educators worry about the diminishing value of evaluating student writing and comprehension skills when AI is so easy and tempting for students to use. 

Another notable chapter in AI’s ubiquity involved a TikTok user, @ghostwriter977, who released an AI-generated track mimicking Drake and The Weeknd (now removed from his TikTok profile). The song, “Heart on my Sleeve,” gained massive popularity, fooling many fans, and raising major concerns and terrifying many in the music industry. For those whose profits are dependent on managing an artist’s image and intellectual property, what impact does this loss of control have on their bottom line when AI replication is now abundantly easy to do. 

AI-driven Scams, Counterfeits, and Theft 

But troubling stories of AI’s misuse keep coming. The latest frightening developments include phishing scams that capture the cadence and speech patterns of loved ones to con people of their money. Others fear of the implications of provocative deep fakes videos of world leaders that could incite global conflicts. We live in a deeply sceptical age. How will deepfakes change people’s trust in the (fake)news. And as writers and actors in Hollywood have taken to the streets in a months-long strike, one major grievance brought to the centre involves AI. Films featuring entirely AI-generated scripts are in the works. Scripts can now be AI-generated. These scripts are modelled on the personal writing characteristics of some of the most celebrated living writers. These writers work has been scraped from the internet, dissected by AI, and then replicated without consent or acknowledgement.  A whole new age of AI-generated scamming, counterfeiting, and theft is on the horizon. This will only accelerate as technology improves and the language models are refined.  The cat is out of the bag.  While governments around the world are trying to legislate to curtail the troubles with AI, it is important to question whether their efforts are too little, too late. 

Ethical use of AI in our studio

These are the topics we’re currently discussing around the office. We wonder if and how AI can be used ethically in a small studio to find clever ways to work better, faster, smarter. When ChatGPT and Dalle-2 first launched we also dove in, experimenting with these tools to see if and how AI could streamline our work, aid in our creative processes, and be the backbone of some truly interesting and engaging future exhibits. The short answer is yes, it can and does, help us each day. However, we acknowledge that like any truly (I cringe to say it) disruptive technology, artificial intelligence can be used maliciously or positively. It depends on the prompts and the motivations of the person in control of said prompts.  We feel the positive application of AI in the design process justifies its use but requires definition to distinguish good use from malicious use.

  • AI (Artificial Intelligence) refers to the simulation of human intelligence processes by machines, enabling them to perform tasks that typically require human cognitive abilities, such as learning from experience, problem-solving, and decision-making. 
  • Generative AI involves the use of algorithms and models to create new and original content, such as images, text, or even music, without direct human input. It generates outputs based on patterns and data it has learned from during its training. 
  • Neural networks are computational models inspired by the structure and functioning of the human brain. They consist of interconnected nodes or “neurons” that process and send information, enabling them to learn from data patterns and make predictions or classifications. 

We use generative AI in the five following beneficial ways: 

#1 AI sparring partner  

One of the best uses for ChatGPT is to spark new ideas. If you have a question or problem, you can drum up a series of prompts for ChatGPT.  The output results are usually unusable as is. But the act of taking the time to write down the problem in prompt form and read the AI-generated response can help you as a concept designer elucidate your own ideas. This often leads to the best solution to the issues.  ChatGPT can act like a best friend who listens to you complain for hours about your problem so you can arrive at the solution on your own.  

#2 AI click monkey 

A digital illustration often involves working for hours on menial, repetitive tasks.  For instance, the visuals needed for a concept of something that hasn’t yet been created. These illustrations are essential to convince a group of stakeholders that a concept is worth investing in. It might take hours if not days to find the perfect balance of elements.  Often a photoshopped concept drawing might be the composite of 20 or more snipped out, position and balanced images needed to capture the essence of a scene.  Each of these pieces has its own copyright statuses that need to be understood and adhered to avoid fines from unauthorized use. It is a lot to navigate.   

The new Adobe Photoshop Beta was recently released for testing. While still in development, it already shows so much potential. It is safe to say it will change design work forever.  The new generative fill tools allow users to quickly add and remove elements, fill backgrounds, iterate, test and prompt entirely new (copyright-free) images all within Photoshop. Without exaggeration, it will save countless hours of boring click monkey work.  It doesn’t replace the illustration process, but it certainly augments and speeds up this process by removing much of the tedium. 

#3 AI prototyping 

Sometimes the biggest obstacle to designing a concept illustration is by imagining how each element can be composited together. This process can take many attempts to get right. But now that generative AI is readily available, this process is much simpler.  The use of text-to-image AI generators such as Dalle-2, Midjourney or Stable Diffusion offers quick ideas of how a composition could look.  If you include all the elements that need to be illustrated together into the same prompt, generate a series of images, you often can get inspiration of the right ways to balance everything into the most persuasive or attractive way.  Again, the results rarely work as generated. But generated results as a quick sketch can guide the creation of an illustration by a skilled human illustrator. 

#4 AI upscaling and frame interpolation   

Pixels in images and frame rates in video are finite by nature. As a result, they are often the limiting factor in many creative processes. Where in previous years, if you wished to use an image in a banner that would span a room, the pixels of the source images determined the quality of the final design.  The same went for video.  If there was a project that needed smooth video playback of 60 frames per second or higher, it all depended on the original material and whether it was recorded in the required framerate or not.  This has changed significantly with the development of open-source AI generative software such as Real-ERSGAN and DAIN. They programs use AI to generate new pixels around existing ones or new video frames in between the recorded ones. Even the latest versions of Photoshop come with upscaling functionality that helps the design process by creating new opportunities to use images that were previously disregarded because they were lacking in pixel density. These tools are easy and quick to use and will continue to improve the workflow for projects, especially as screens increase in resolution and quality and rate frame requirements continue to increase.  

#5 AI game design 

Unity has developed AI tools to accelerate content creation and aid in generating draft assets. They have introduced two AI products: Unity Muse, a platform for AI-driven aid during creation, and Unity Sentis, which allows embedding neural networks in builds for real-time experiences across multiple platforms. We’re busy experimenting with these tools to speed up our game design process and see what interesting concepts come from these tools. 

The paradoxes  

Artificial intelligence can be beneficial by democratizing creativity, giving individuals or small teams of people access to tremendously powerful tools. This will lead to a new era of creative works as indie studios produce content that was normally limited to major studios with enormous budgets.  

The problem is these tools are not being used solely by indie studios.  Major studios see the potential of AI to drive profits, cut costs and keep their hegemony. What may happen is fewer designers, game artists, writers, and voice actors being hired in big budget projects, because in the end, these creative jobs are expensive. Now much of the creative process can now be outsourced to AI workforces. This is ironic because for decades as discussions about human labour being replaced by robots and AI took place, it was argued that the creative sector was safe from automation. It was assumed that creativity couldn’t be replicated by machines.  We are currently witnessing the limits of this assumption. 

It’s too soon to say. Afterall, at the centres of ChatGPT, Bing and Bard are a mechanism finding patterns in human communication and ideas that have already taken place in the past. They are transformations based on what has existed previously. Can these really be considered new ideas?  Likewise, Dalle-2, Midjourney and Stable Diffusion are modelled on vast oceans of existing image data. They are built on human creativity of the past, like distorted echoes. Maybe there will always be the need for original creations from human minds. Maybe we’ll value this more in the years to come in some pushback AI, the latest shiny new machine. Time will tell. 

Our in-house artist Dirk Jan made this image using his own skills and AI. Read more about it here