Imagine you hand a box of crayons to a second grader and ask them to draw a house in a field. They don’t just rip out a page from a coloring book or copy a picture they’ve seen before. Instead, they use their understanding of what a house and a field look like to create something new. They might draw a big red house with a blue roof and surround it with a green field dotted with yellow flowers. Each element—the house and the field—comes from the child’s memory and imagination, mixed together in a unique way.

This simple analogy is a great way to start understanding how artificial intelligence (AI) generates new images, like if you asked an AI to create a picture of a house in a field. The AI isn’t pulling this image from a database or copying it from somewhere else. Instead, it creates something completely new, much like our second grader.

A recent ad run by Adobe that stated “Generative AI – This Changes Everything!” invoked a ton of negative comments from users. They all fit into 2 categories:

  • Adobe is stealing artwork from people
  • AI is going to take away the jobs of artists

That’s just a few of the thousands of comments and hate messages Adobe is getting. I certainly get it – when I started advocating AI, I got plenty of hate mail as well, with comments like these:

“You’re a ******** moron! Your webinar probably caused a lot of people to lose their careers.”

“AI is just plagiarizing other content. My husband is an artist so I am super sensitive to people stealing his work.”

The reality of how AI works is this:

Training: Initially, the AI goes through a phase called training, where it is exposed to numerous images, learning from each. This includes many houses and fields, but it’s not memorizing these images. It’s learning what characteristics make up a house (like windows, doors, and roofs) and what elements define a field (like grass, paths, and trees).

Feature Learning: During training, the AI develops an understanding of these features independently of the specific images. It learns various styles of roofs, different types of windows, and numerous ways grass can look in different lights and settings.

Generating New Images: When tasked to generate an image of a house in a field, the AI uses this learned information to create a new image from scratch. It synthesizes features in novel combinations—perhaps creating a house style that hasn’t been seen before, placed uniquely in a field setting that doesn’t exactly replicate any single image it has seen during training.

Manipulating and Combining Features: The AI adjusts these features, manipulating them based on the vast amounts of data it has processed, to create a cohesive image that makes sense to human viewers. For example, it might generate a rustic wooden house surrounded by a lush green field under a sunset sky, combining elements in a way that is both fresh and realistic.

Evaluation and Refinement: Advanced AI models, particularly those using techniques like Generative Adversarial Networks (GANs), even have mechanisms to critique their work. A part of the AI acts as a critic, assessing if the generated image looks realistic enough. If not, it sends this feedback to the generative component, refining its approach until the output is improved.

Just like with images, AI can also generate new text and music. For text, AI models learn from a vast database of written content, grasping grammar, style, and different forms of narrative to create original essays, stories, or poetry that mimic human creativity but are entirely original. In music, AI analyzes patterns in rhythm, melody, and harmony from various genres to compose new pieces that might sound familiar but are genuinely unique compositions.

This process shows that AI is capable of true creativity, using learned data to produce new, original works. It’s not about copying; it’s about understanding and innovating. Just like our second grader, AI takes the essence of things it has seen to create something entirely new, not previously existing. This capability makes AI an incredibly powerful tool in fields such as art and design, where generating novel ideas is key. So, when we see an AI-generated image, we’re not looking at a simple collage of pasted-together bits from its training data, but at a piece of original art, created through a complex interplay of learned concepts and innovative synthesis.

While there are concerns that AI might replace human jobs, especially in creative fields, it’s important to view AI as a tool rather than a replacement. For those who embrace it, AI can significantly enhance human creativity, opening up new possibilities and allowing artists, writers, and musicians to push boundaries beyond traditional limits. By automating certain aspects of the creative process, AI frees up human artists to focus on the bigger picture—conceptualizing and refining ideas that are uniquely human.

Further, AI can assist in areas where there’s a lack of resources or skills. For example, small businesses or independent artists can use AI to generate promotional materials, design products, or even compose music for projects, which might otherwise require expensive professional services. In education, AI can help tailor learning experiences that foster creativity in students at a personalized level.

Ultimately, rather than having a scarcity mindset that sees AI as a threat to jobs, it’s more fruitful to adopt a mindset of abundance. AI is here to amplify our creative potential and help us succeed beyond what we previously imagined possible. By leveraging AI as a collaborator, creative individuals and professionals can explore new artistic landscapes and achieve greater heights in their careers.