webneel

The Relationship Between Art and AI

Grab popcorn if you see an artist and an AI expert getting into a war of words. "Nah! Machines can't replace the ethereal reins of artisans." "Ohhh, well! You better know our machines. They can predict your thinking patterns and express art as brilliant as you." "WHAT! You must be kidding me!! Art without soul and artists without emotions can't have a transcending finesse." "Sorry, but we can train machines for artistic expression more than ever." If you have been a part of such debates… congratulations!!! You achieved nothing!

Because...Like you can't separate science and art; you can't separate AI and art!

They can exist in a common plane without having to be dismissed by the other.

The real meaning of it lies in understanding AI is not an enemy with artists or any human, but their friend!

The idea of having an AI machine was to ease artist's work rather than to compete with them.

See how Jeremy puts AI into the painting perspective.

Take a look at Jeremy Krabil.

He couldn't paint.

But he could design an AI-based software that could paint for him.

Would Jeremy be less of an artist here?

Or do we call his machine the originator of the piece?

Krabil sold his first AI artwork named "Edmond de Belamy" for over $430,000.

The AI robot was required to learn what it took to be "creative."

But Krabil programmed the machine to be artistic.

So was painting the output of its designer?

Well, that's your call to decide!

But you can't separate AI with arts and crafts now!

So this is how AI makes a fantastic artist...

AI puts two intelligent network machines into conflict and lets 'em create something new for the observer!

To make art happen, AI uses a generative adversarial network (GAN).

Now what exactly is a GAN?

GANs are nothing but algorithmic architectures.

So the process of AI art designing looks like this:

GAN takes two neural networks in its architecture.

One network machine is called a generator, and the other is termed a discriminator.

Then these two networks are pitted against each other (like a conflict).

You can think of them, more or less, like two intelligent machines working against each other.

Then you feed an extensive data set to the GAN.

The generator then picks these data sets and engenders random sounds, images, or objects (what you want them to make).

The discriminator picks the outputs generated by the generator and matches them with artistic images (already existing in the data set).

So the discriminator basically picks the learning sources from the precedent, mix-matches the output with the existing data, and then refurbishes the data into its own random expression.

Finally, what you have in the end is the ultimate synthesized piece— be it in the form of image, video, or voice.

Isn't it how our brains work?

Don't we follow the same algorithm?

We map "the more creative" right with "the more analytical" left?

So can you say, "humans are machines, and machines are human?"

Well, we'll leave it for the philosophers to decide.

And then some machines can create art without the programmer's input…

If we can't discipline art and AI in the same plane...
and if there's to be a "huge" difference between the artists and the machines...

Should it be an easy task for everyone to distinguish between paintings created by both?

I think so. But it doesn't seem to happen!

The "bot artists" at Rutgers Art and Artificial Intelligence Lab (AAIL) are the "living" testimonies.

They have been trained hard to create paintings that could fool anyone into believing the art is human-designed.