VISIONS TO VISUALS
Gen AI puts creative expression within everyone's reach.
Gray Tabby and Granola, oil on canvas, Imagen 3
Text to image generative AIs are designed to communicate with humans using the "natural language," or the everyday language we use to communicate with one another. It opens up new creative opportunities to all human users without having to learn a special language to communicate with a machine (a.k.a. "coding").
And yet, generating images with AI tools is not a cookie cutter process. To turn our visions into visuals, human users must learn how to verbalize and communicate their ideas to AI. Since human intelligence and artificial intelligence don't work in the same way, what seems "obvious" or "natural" to humans may not be so to AI; or AI might "read into" something you mentioned and take the image in the direction you didn't intend. While these differences can make the collaboration with AI a rather frustrating experience, especially in the beginning, they also make the moment of success all the more exhilarating.
Storytelling with AI
Storytelling is the art of conveying a message or idea through a narrative that engages the audience emotionally and intellectual. This form of communication, practiced among diverse human communities across time and space, is equally important in digitalized contemporary society of ours, as an effective vehicle to reach out to our audience especially when our communication is "mediated" - i.e., through technological means lacking direct human contact - and suffers from the common challenge of disconnection.
Until recently, contemporary storytelling was a complicated endeavor of the professionals, who brought specialized knowledge and skills to decide what story to be told (advertising executives), to write a narrative (script writers), to design the visual appearance (graphic designer, director), and to enact the story as imagined (anime technicians, actors).
Digital technology, the Internet, and now, expanded accessibility of ever more sophisticated and user-friendly AI models has changed the landscape entirely. Now, any ordinary person with a story in their heart can leverage the assistance of generative AI to write their own narrative, create images or video clips, and deliver their story through a social media platform, YouTube, Instagram, Tik Tok, you name it!
Training Data Question - Latest News
One of the arguments against using AI image (and now video) generators is the ambiguity around the use of human-created art to train AI. When genAI models are trained on visual images - photographs, paintings, drawings, illustrations, etc. - obtained without compensation to human artists and graphic designers, does it constitute a kind of digital "theft" and an infringement of creators' rights over their artistic expression and intellectual properties? ​​
​
It's difficult to wrap our heads around how AI is trained and how it "generates" text and images, and it is understandable that many people who saw the images AI produced assumed that it was "copying" images it has stored. Earlier concerns and opposition was, in large part, the result of this misconception about how human-created visual materials are used by AI.
Recent court case in UK is considered a "win" for AI, in which the court rejected Getty's claims that Stable AI, infringed upon its intellectual property rights when it used millions of images from Getty-owned images for the training of its popular image generator Standard Diffusion, because the AI model did not "store or reproduce any Copyright Works (and has never done so)." ​​
This is a complex issue and no one court case will clear up the concerns of creators and others who are concerned with intellectual property rights. However, it is important to note that the clear distinction was made by the court between an AI "learning from" visual materials vs. "copying and reproducing" them.