Did a filmmaker do enough to ensure the accuracy of AI-generated content?
Here’s the situation:
Paleontologists at a local university in Utah recently unearthed new dinosaur fossils. Inspired by the discovery, as well as her child’s fascination with these creatures, filmmaker Tina Rexford decided to make a documentary series about ancient dinosaurs that roamed the western United States, which would conclude with the latest discovery.
While she was excited about the project, Rexford was concerned about how to get three hours’ worth of visually engaging film on a tight budget. She had recently learned about new artificial intelligence (AI) tools that would generate images and audio from text prompts. They were inexpensive and easy to use, so she decided to give that a try.
The filmmaker was amazed at how realistic the photographs were and how easy they were to modify to suit the storytelling. With a few simple prompts, she was able to generate an opening scene with a T. Rex attacking a gigantic Apatosaurus. As the battle ensues, viewers could hear the roars of the dinosaurs. All of the visuals and audio were created using generative AI, but the producer decided it would be too distracting to try to tell the audience. The scene obviously took place in pre-historic times, so there couldn’t possibly have been real pictures or audio.
The producer uses AI again later in the series to show the remains of the Apatosaurus being buried by volcanic ash and then a timelapse scene of the bones becoming fossils. During this timelapse, viewers see Native Americans wearing feathered headdresses as they traverse the landscape over the buried remains. The series concludes with authentic contemporary footage of the paleontologists discovering the fossils.