How Ai Can Replace Visual Artists
Here is a controversial thought: artificial intelligence can create beautiful paintings just as well as humans can.
The image above is Pierre Fautrel posing with his portrait of Edmond de Belamy, the first Ai-generated artwork that sold for $432,500 at auction. Okay — probably not as good — but I’m sure Picasso wasn’t that good when he started either.
Considering Ai paintings will start to become as advanced and complex as human art, does this render the future visual artist obsolete?
Artificial intelligence is now at a stage that has turned distant dystopian ideas into a reality: driverless cars, facial recognition, automated legal services and so on. Our future is less likely to be about driving taxis or reading legal documentation, but somewhat more uncertain.
“Automation will displace between 400 and 800 million jobs by 2030”
In contrast to many other industries, the creative and cultural industries might seem safe from mass job displacement. But the question is: can you fully automate creativity?
An attempt to answer this takes me back to how Edmond de Belamy, the AI-generated Portrait, was created.
The team behind the portrait: the collective Obvious, needed quite a few ingredients for the machine learning process to create the Edmond de Belamy portrait. I’ve attempted to outline the process below:
- Collecting a dataset consisting of thousands of portrait images. To be precise, they collected a total of gasp 15,000 original portraits painted between the 14th and 20th century.
- Cleaning and splitting the dataset into training and validation data, while removing any outliers.
- The first ‘Generator’ computer model then creates images based on patterns seen in the training dataset.
- These images are passed to the ‘Discriminator’ computer model, which then determines if the portrait images are human-made by: comparing the validation dataset of human-made portraits to the computer-made ‘Generator’ portraits
- The Generator then takes the feedback from the Discriminator and updates its model. We repeat the process until the Generator has a good-enough model whose generated images successfully fools the Discriminator into thinking that the portrait images it produced are human-made
- Once this happens successfully a computer-portrait portrait has passed the fake or not-fake test — this is how the Portrait Edmond de Belamy was created
So the team made meticulous efforts to firstly acquire these ingredients, these being the 15,000 original portraits painted between the 14th and 20th century, then they proceeded with an equally meticulous process of perfecting the machine-learning model to create the final artistic product.
Does it still mean that Ai created the art then? Arguably not, since the models were created by human artists. The Ai-generated Portrait of Edmond de Belamy can be appreciated visually, sure, but only because the computer paints by examining thousands of original artwork images to produce something ‘original’. Essentially all the machine learning model did was simply create an aggregate of the portraits it had learned from.
Alan Turing, computer scientist and philosopher, argued that this type of information processing is precisely what humans do, and creativity is merely an output from many inputs.
He argued that machines are able to think creatively in a similar way to humans. However, contemporary philosophical thinkers have some valid counter-arguments.
Carl Frey, author of The Technology Trap, makes a distinction between the boundary of ‘human’ and ‘non-human’ artist creativity:
“Artists don’t just draw upon pre-existing artworks, they draw upon experiences from all walks of life — maybe even a dream — and a lot of our experiences are always going to be outside of the training dataset.”
So maybe we don’t need to worry yet — our visual artists are safe from being replaced with silicon chips — for now! Computer creativity isn’t exactly going to replace human creativity anytime soon.
As Michael Wooldridge, professor of computer sciences at Oxford university, puts it: people will have to wait a long time for artworks created by Ai to “deeply engage” them.
References:
- https://www.iotforall.com/impact-of-artificial-intelligence-job-losses
- https://www.ft.com/content/e082b01d-fbd6-4ea5-a0d2-05bc5ad7176c
Further reading:
- https://www.artsy.net/article/artsy-editorial-art-failing-grasp-christies-ai-portrait-coup
- https://www.theverge.com/2018/10/23/18013190/ai-art-portrait-auction-christies-belamy-obvious-robbie-barrat-gans
- https://www.christies.com/features/A-collaboration-between-two-artists-one-human-one-a-machine-9332-1.aspx
- https://medium.com/swlh/can-machines-think-9fc81e61ac6