ReadyAI and the Transformation of Creativity: Charting the Course of AI-Driven Artistry in K-12 Education
By: Rooz Aliabadi, Ph.D.
When we think of Artificial Intelligence (AI), the thoughts that come to mind usually revolve around advanced robots, cutting-edge technology, or complex algorithms. Rarely do we consider its incredible application in the creative world. Today, we’re taking you on a journey into the captivating realm of AI in art, tracing its roots and exploring its exciting potential for today’s and tomorrow’s classrooms.
Believe it or not, the concept of AI in art isn’t as modern as it may seem. It dates back several decades, to the early 1970s, when an artist named Harold Cohen challenged the boundaries of creativity and technology. Cohen taught an early AI system called “Aaron” to instruct a robot to sketch black-and-white shapes onto paper. Over the next decade, Cohen’s AI creation was not only drawing simple shapes but had progressed to drawing human figures. The concept was revolutionary, demonstrating how the technology of the time could be harnessed in previously unthinkable ways.
In today’s digital era, the concept of AI in art has expanded tremendously, with “generative AI” models becoming increasingly prevalent. These intelligent models can put a digital brush to a virtual canvas, creating impressive images in seconds based on user-provided text prompts. This groundbreaking technology is accessible to the public through applications like Midjourney and OpenAI’s DALL-E. These AI-generated creations are incredibly realistic, often fooling human observers into believing they are looking at manually crafted art. A memorable instance was when AI-generated images depicting Donald Trump being handcuffed by police went viral online. But how exactly do these advanced AI models operate, and what steps are being taken to refine their capabilities?
At the core of generative AI models lies an intricate technology known as deep learning. Drawing inspiration from the human brain’s neural network, deep understanding uses layers of interconnected nodes. These models are not just fascinating in their structural design but also their extensive training process. They are fed massive databases of images, such as LAION-5B, one of the largest public databases containing an astounding 5.85 billion tagged photos. These images are from various online sources, including social media platforms, stock-photo libraries, and e-commerce websites.
The most advanced image generators use a specialized generative AI called a diffusion model. Diffusion models add a distorting visual “noise” to images, obscuring the pictures until they become unrecognizable. However, by learning to reverse this process, the AI can reproduce an image that closely mirrors the original. This intricate procedure enables the AI to identify groups of pixels corresponding to particular visual elements, which it then compresses, categorizes, and stores in a unique mathematical cache called the “latent space.”
To illustrate this, imagine you instruct an AI generator to create an image of a hippopotamus. The AI, having learned which types of pixel combinations correlate with the word “hippopotamus,” samples from its latent space to generate a believable image of the creature. Should you provide a more detailed prompt such as “a Renaissance-era oil painting of a green hippopotamus by the River Nile,” the model must delve deeper into its latent space. It would need to source additional layers of visual detail such as image style, texture, color, and setting and then skillfully combine them to meet the specific request.
Despite the impressive capabilities of AI image generators, they sometimes produce unusual results, mainly when dealing with complex or ambiguously worded prompts. Scenes underrepresented in their training data can also lead to unpredictable outputs. Common hiccups depict hands with abnormal proportions or extra fingers, facial asymmetries, or text containing non-existent letters or gibberish words. But these are all part of the learning process. Developers continuously refine these AI models by improving their training data and adjusting their algorithms. Recent enhancements, such as those made by Midjourney to strengthen its hand generation, show how swiftly the field advances.
Looking ahead, the future of AI in art promises to be exciting. It may soon become a Herculean task to distinguish between an AI-generated image and an actual photograph or painting. Just as AI is transforming the art world, so is the realm of education. As we incorporate these advanced technologies into our K-12 classrooms, we embark on a thrilling journey where creativity meets technology. So, let’s buckle up and delve into this mesmerizing world of AI-generated artistry — an exploration guaranteed to enrich our educational experiences!
Among the pioneers leading the way in this AI-artistry education initiative is ReadyAI. We have been diligently crafting comprehensive lesson plans for the K-12 audience on image generation. These lesson plans offer a unique opportunity to integrate this forward-thinking technology into the curriculum, giving students a hands-on experience with AI-generated art.
ReadyAI’s curricula provide theoretical knowledge and encourage practical application, fostering a creative learning environment that blurs the lines between art and technology. Students can delve into AI image generation, learning how to instruct AI to create fascinating images based on their prompts. They understand the concept of ‘latent space’ and ‘deep learning,’ which are fundamental to AI image generation.
As teachers, we can ignite curiosity and cultivate knowledge in transformative ways. Bringing AI artistry into your classrooms not only equips your students with a nuanced understanding of this emerging technology but also encourages them to think creatively, solve problems, and learn from mistakes — essential skills in our rapidly evolving digital world. So why wait? Let’s take the leap, embrace the exciting opportunities ReadyAI offers, and empower our students with the tools to shape the future of art and technology!
This article was written by Rooz Aliabadi, Ph.D. (rooz@readyai.org). Rooz is the CEO (Chief Troublemaker) at ReadyAI.org
To learn more about ReadyAI, visit www.readyai.org or email us at info@readyai.org
