We are the artist: generative AI and the future of art

Before writing a single word of this article, I created the above image using a new type of AI software that produces “generative illustrations.” The process took about 15 minutes and no paint or canvases were involved. I just entered a few lines of text to describe the image I wanted – a robot holding a paintbrush and standing by an easel.

After a few iterations, tweaks and revisions, I achieved a result I was happy with. To me, the image above is an impressive piece of original artwork. After all, it appeals to the imagination and evokes an emotional response that seems no less authentic than human art.

Does this mean that AI is now just as creative and evocative as human artists?

No.

Generative AI systems are not creative at all. In fact, they lack any real intelligence. Sure, I typed in a request for an image from a robot holding a brush, but the AI ​​system had no real understanding of what a “robot” or a “brush” actually is. It created the artwork using a complex statistical process that correlates images with the words and phrases in the prompt.

The results look like human works of art because the system has been trained on millions of human artifacts — drawings, paintings, prints, photographs — most likely captured from the Internet. I’m not saying that these systems aren’t impressive. The technology is truly amazing and very useful. It’s just not “creative” in the same way people think of creativity.

After all, the AI ​​system didn’t do that to feel everything while making the work. It also failed to take into account the emotional response it hoped to evoke in the viewer. It did not draw on any inherent artistic sensibilities. Essentially it did nothing that a human artist would do. Still, it did a remarkable job.

The image below is another example of a robot holding a brush generated during my 15 minute session. Although it was not chosen to be used at the top of this article, I find it very compelling work, imbued with unmistakable feeling:

If the AI ​​isn’t the artist, then who is?

If we consider the above pieces to be original works of art, who was the artist? I certainly wasn’t. All I did was enter a text prompt and make a variety of choices and revisions. At best, I was an employee. Nor was the artist the software, which has no understanding of what it has created and no ability to think or feel. So, who was the artist?

My opinion is that we all made the work of art – humanity itself.

I believe we should regard humanity as the artist of the record. I don’t just mean people who are still alive today, but everyone who has contributed to the millions of creative artifacts that generative AI systems have been trained on.

It is not only the countless human artists who have had their original works vacuumed and digested by these AI systems, but also members of the public who have shared the artwork, described it via social media posts or simply upvoted it so that it became more prominent in the huge database we call the Internet.

To support this idea, I ask you to imagine an identical AI technology on a distant planet, developed by another intelligent species and trained on millions of their creative artifacts. The output of that system can be artistic for them – suggestive and impactful. It would probably be incomprehensible to us. I doubt we would recognize it as art.

In other words, without being trained on a database of humanity’s creative artifacts, current AI systems would not generate anything that we would recognize as emotional artwork. Hence my claim that humanity should be the artist of the record for large-scale generative art.

A fee

If an individual artist took the robot photos above, they would be compensated. Likewise, if a team of artists had created the work, they too would be compensated. Big budget movies are often staffed by hundreds of artists in many disciplines, all contributing to one piece of art, all compensated. But what about generative artwork created by AI systems trained on millions upon millions of creative human artifacts?

If we accept that humanity is the artist, who should be compensated? It is clear that the companies that provide generative AI software and computing power are earning significant compensation. I don’t regret paying the subscription fee it took to create the artwork above. But there were also huge numbers of people who took part in the creation of that artwork, their contributions inherent in the vast array of original content on which the AI ​​system was trained.

Should humanity be compensated?

I believe it is reasonable to consider a ‘humanity load’ on generative systems trained on huge data sets of human artifacts. It may be a modest transaction fee, perhaps paid into a central “human fund” or distributed to decentralized accounts using blockchain.

I know this might be a strange idea, but think about it this way: if a spaceship full of enterprising aliens showed up and asked humanity to contribute our collective works to a massive database so they could use derived human artifacts for profit. we would probably ask for compensation.

Well, this is already happening here on Earth. Without being asked for permission, we humans contributed a huge collection of creative artifacts to some of the greatest companies this planet has ever seen – companies that can now build and operate generative AI systems. to sell derivative content for a profit.

This suggests that a “humanity tax” is not a crazy idea, but rather a reasonable first step in a world that is likely to use more and more generative AI tools in the coming years. Our contributions won’t just be used to create quick images at the top of articles like this one. Generative methods will be used for everything from creating written essays and blog posts to generating custom videos, music, fashion and furniture, even beautiful artwork that you hang on your walls. All this will draw on large parts of the collective works of humanity – the artist of the record.

Louis Rosenberg, Ph.D. is a pioneer in VR, AR and AI. His work began more than thirty years ago in Stanford and NASA labs.