Author | 99-pound Dumbbell
Image | Internet
First Reading | For reprints, please contact for authorization
“Do you know about AI painting?”
“How could I not know? It’s been really popular online lately.”
Image source: Wuqing Xiaoliang Weibo
“This AI is not the same as that AI.”
This AI refers to Artificial Intelligence, not Adobe Illustrator
Recently, AI painting has become a hot topic online. Not only are professional artists from various industries discussing it, but ordinary people are also highly engaged in the conversation. What exactly is AI painting? Does it have anything to do with us?
What is AI Painting?
You may have seen AI illustrations online; they can transform a photo into a beautiful comic in seconds. Some AI paintings are stunning, leading many to wonder, is this really AI-generated?
Works created by AI
AI painting is a method of creating images automatically generated by intelligent computers. We can use artificial intelligence algorithms to create art, which simply means artistic creation completed through algorithms.
The working principle is that based on a vast collection of existing human images, users can classify and recognize them through algorithms, and then generate new images. The entire process can take as little as a few seconds.
AI Painting is Actually Older Than You Think!
Many people view AI painting as a new trend. However, this is not entirely true.
AI painting began much earlier than we might imagine.
As early as 1973, artist Harold Cohen started experimenting with a computer program called AARON for artistic creation. This is different from today’s AI painting, as AARON controlled a robotic arm that painted on a canvas, unlike current AI that generates images through algorithms.
Harold Cohen and his AARON
Although AARON created many works during this process, it learned through imitation and reference, and still struggled to reach a level where it could create a unique style. Moreover, at that time, AI was also creating based on the style of the experimenter, marking the embryonic stage of AI painting development.
AARON’s artwork created in 2004
Later, with the emergence of the Python programming language, AI painting began to utilize programming for image creation.
Following the previously mentioned principles of AI painting, it wasn’t until 2012 that Andrew Ng and Google senior expert Jeff Dean used 16,000 CPUs from Google Brain to train a large neural network to generate a blurry image of a cat’s face.
Andrew Ng and the cat face generated by Google Brain’s deep learning
Reportedly, Ng and his team trained for three days using photos of cat faces from 10 million YouTube videos to finally obtain the cat face image. Although this training was inefficient compared to today, it marked a breakthrough in the field of computer vision, paving the way for the AI painting we discuss today.
Quick Fact: Andrew Ng was a professor and director of the Artificial Intelligence Laboratory at Stanford University, considered one of the leading scholars in the field of artificial intelligence. In 2010, Ng joined Google to lead the establishment of the renowned Google Brain project, aimed at deepening Silicon Valley’s understanding of deep learning.
After Ng’s cat face model emerged, academic research on AI painting reached unprecedented progress.
For instance, in 2014, the initial image generation model GAN (Generative Adversarial Network) was developed, consisting of two parts: a generator and a discriminator, where the former generates images and the latter assesses them. This image generation model greatly propelled the development of AI painting.
Until April 2017, when Google proposed the Sketch-RNN model, it gained widespread public attention, leading to a surge of drawing models. In July of the same year, Facebook collaborated with Rutgers University and the College of Charleston’s Department of Art History to develop the Creative Adversarial Networks (CAN) model, allowing AI to attempt drawing in different styles and images.
Artwork created by the Creative Adversarial Network CAN1
Interestingly, researchers organized a test asking people to guess whether these artworks were created by artists or AI.
The result showed that 53% of people believed that the AI artworks from the CAN model were created by artists, marking a historical first where over half of respondents thought so.
Looking back at 2022, numerous AI painting tools flooded the market, significantly lowering the development threshold for AI drawing applications and pushing the topic and concept of AI painting to new heights.
AI Painting Also Has Another Voice
Questions? Disapproval?
When the public was not yet aware that AI could paint, AI painting gradually attracted attention amidst novelty and skepticism.
Space Opera
This piece, “Space Opera,” was generated using the AI drawing tool Midjourner, then color-adjusted in Photoshop.
Despite the public’s criticism, the judges stated they would still award the championship to this painting. The author remarked, “I won’t apologize; I won. I didn’t break any rules.”
This inevitably reminds one of the impact of the invention of the camera on the art world at the time. Over time, photography has become an art form on par with painting, without replacing it. Now, the popularity of AI painting raises a question:
In fact, AI has already penetrated our lives beyond just painting.
Your smart speaker sometimes says, “Looks like I’m not smart enough; I need to keep learning,” which is not just a saying; it genuinely analyzes and learns from big data: the voice assistant is also powered by AI technology. Microsoft’s Xiaoice even published her own poetry collection;
AI can also compose music; with current technology, AI can reach the level of composing general pop music;
There’s also AI facial recognition, AI one-click poster generation, AI intelligent photo editing…
When AI simulates human work,
we feel both surprised and concerned.
The surprise is that AI is developing so rapidly
The concern is,
In which fields will AI gradually replace humans?
What are your thoughts?
Scan to follow
First Reading
Reading · Accompanying · Growing
