In the evolving landscape of technology, AI has emerged as a creative force that is reshaping the way people learn and work, but questions have risen regarding the ethical use of AI-generated content.
AI tools ChatGPT and Stable Diffusion were the primary topics of professor Bruce Hutchinson’s keystone conversation Nov. 9.
Hutchinson, the film program coordinator at UCA, described himself as somewhere in the middle of the debate on just how significant the role AI plays in people’s day-to-day lives.
“I call myself an interested observer, someone who wants to use AI as a tool,” he said. “I don’t think AI is going to take over and destroy the world as we know it. I also don’t think that AI is going to save our lives and make us live in a glorious utopia.”
Hutchinson then clarified misconceptions about what AI actually is, versus the commonly used tools we describe as AI.
“What we all acknowledge as AI tools are not really AI tools at all, in terms of the classical definition of artificial intelligence. These are just more advanced technological computer tools that we make use of. So the kind of tools I’m talking about are tools like ChatGPT,” Hutchinson said.
Hutchinson shared common uses of ChatGPT, including citing sources and even writing recommendation letters. Then, Hutchinson expressed his desire for more innovative uses of the text-based AI generator.
“I want to explore more creative uses of ChatGPT and how it can be used to foster our creativity. For example, like giving us ideas or leading us down a path to come up with ideas that we wouldn’t come up with otherwise,” Hutchinson said.
Hutchinson then offered a more in-depth analysis of what exactly ChatGPT is and how it functions.
“ChatGPT is made by a company called OpenAI. It is a text-based AI and it simply generates text, but how it generates the text is more interesting. It uses what is called a deep learning model, and basically, these models are fed huge amounts of data from the internet. Then they learn to recognize and label all of that information,” Hutchinson said.
Hutchinson used Stable Diffusion, an image-based AI also made by OpenAI, as an example.
“If you give Stable Diffusion a million pictures of a dog, and then you tell it to create a dog, it kind of knows what a dog is because it has seen millions of pictures of dogs,” he said. “Then the more parameters we give it, the more detailed it gets. If you tell it to create a dog in the park, it can, because it has also seen millions of images of parks. It puts the images together and creates a dog in the park.”
Hutchinson discussed the ethical implications of using these text-based and image-based tools.
“So now I want to explore the use of primarily public data sets that are used to train all of the AI to do the things it does. So what does that mean about how it creates that information, how it makes those images and who owns all of it? Well it’s all of our stuff, it’s all work humans created.”
Hutchinson continued, “So if we have AI using copyrighted images, movies, songs, books and information to generate potential answers and ideas for us, is that ethical?
Hutchinson then provided an answer to the difficult question.
“We as humans obviously get ideas from other places, other works. I think a big part of the ethical question is answered by what we do with the information we get and how closely it resembles the original work.”
