Originally shared by Yonatan Zunger
On the heels of AlphaGo’s stunning victory over Lee Sedol, Demis Hassabis’ (founder of DeepMind) statements about his interest in music and assistance, and things like the pictures produced below by Ostagram, a site created using “Inceptionist” AI to combine images to form new ones, I’ve heard several people lament that artists will soon be obsolete as well. But I think that this misinterprets the nature of AI.
You can think of an art-generating AI as sitting along a spectrum from medium, to collaborator, to fellow artist. All of these are real things, but none of them have obviated the need for artists – rather, they’ve created new opportunities and new methods for art. At one end of the spectrum, these AI tools are media: you can use them to perform unexpected but quasi-mechanical actions, eg using Ostagram’s system to combine two images you made or chose to form something new. (NB that this system takes the objects from one image and tries to render them in the style of the other; a rich source of ideas) Further along, the AI may be suggesting, responding to your choices, interacting with you – that is, working together to create a work that neither of you would have made on your own. And at another extreme, an AI may be making art entirely on its own.
All of these things have happened several times before. Photography did not obsolete painting, but it did have a profound influence on the visual arts, create new kinds of artist, and also created entirely different things like “photojournalists” and “selfies.” Photoshop, even with tremendously sophisticated filters, likewise didn’t obsolete photography. It was certainly used to create junk, but as people learned to use it as a scalpel instead of an axe, it became a critical part of many artists’ toolboxes.
I think this is applicable to many fields where the objective isn’t to create or do one specific thing, but to engage in a dialogue with others. When the interaction itself is the fundamental thing, there’s no sense in which the computer “replaces” it; instead, it participates.
From a business perspective, it’s a little more complicated, but again not trivial. One can think of AI artists as more artists on the market, and perhaps there will be AI-created products which replace things that people use artists for today. (Much like demand for oil portraiture plummeted with the rise of cameras) But the work of two artists is not a substitutable good in most cases. The work created by AIs will often differ from the work of humans not in technical possibility, but in the choice of what to create: an AI would not create the same thing as a human, not because either couldn’t, but because the human and the AI are trying to communicate two different things – just like two human artists.
So I, for one, am not afraid that art will vanish or become the domain of computers alone. Rather, I’m looking forward to the extraordinary ideas of the next few decades, from both human and artificially intelligent creators.