How did AI painting take off? From history to technological breakthroughs, read the history of the hot AI painting in one article

One reason this model has received so much attention is that Google has made the source code open source, so third-party developers can develop interesting AI sketch applications based on the model. One online application is called "Draw Together with a Neural Network", where the AI can automatically complete the drawing for you with a few random strokes.

It is worth noting that the Internet majors have been the main force in the research process of AI drawing models, except for the above-mentioned research done by Google, and the more famous one is the new model obtained in July 2017 by Facebook in collaboration with Rutgers University and the Art History Department of the College of Charleston, called Creative Adversarial Networks (CAN). Adversarial Networks)

As you can see from the portfolio below, this Creative Adversarial Network CAN is trying to output drawings that look like the artist's work, which are unique and not copies of existing artworks.

The creativity in the works generated by the CAN model struck the development researchers at the time, as they looked very similar to the abstract paintings popular in the art world. So the researchers organized a Turing test, asking the audience to guess whether the works were the work of human artists or the creations of artificial intelligence.

As a result, 53 percent of the audience thought the CAN model AI artwork was by human hands, the first time in history that a similar Turing test had broken the halfway mark.

But the CAN AI painting, limited to some abstract expressions, and in terms of artistry scores, is still far from the level of human masters.

Not to mention the creation of realistic or figurative paintings, which do not exist.

In fact, until the beginning of 2021, when OpenAI released the popular DALL-E system, its AI painting level was only average, and the following is the result of a fox drawn by DALL-E, which is barely recognizable.

But it's worth noting that with DALL-E, the AI starts to have an important ability, which is to follow the text input prompt!

Next, let's move on to the question posed at the beginning of this article. I don't know if you feel the same way, but since this year, the level of AI drawing has suddenly risen dramatically, and the quality of the previous work has taken a fundamental leap forward, and it feels like a long time ago.

There must be a demon. What's going on? Let's talk about it slowly.

Why AI painting has improved dramatically

In many science fiction movies or dramas, there is often a scene where the main character speaks to a particularly sci-fi computer AI, which then generates a 3D image that is presented in front of the main character in VR/AR/holographic projection.