This machine paints Jewish avocados

Parents: Don’t let your children take out student loans to be studio artists.

That’s because computers, which already have proven their ability to beat humans at playing chess and delivering junk mail, now can produce custom oil paintings — or photographs, or pictures in the style of your favorite Chinese dynasty— simply in response to a description.

The algorithm underlying this ability, called generative adversarial networks, was published in 2014. But it’s not just math at work — it’s also a database of most of the billions of images available on the internet for a hungry machine to download.

That means these artificial artists require both smart programmers to tweak the algorithms, and deep pockets to afford to store all the masterworks of human civilization, as well as the far more numerous lesser works, including bad memes. (But not too terribly bad memes — the companies behind this are trying to keep their art-generating golems from learning from the worst sewers of the internet.)

A few months ago, the site neuralblender.com began offering access to last year’s version of this technology, which, like a child with shaky eye-hand coordination, does better at surrealism than realism. (This picture of a “Martian synagogue” was made using that program.)

But the revolution came in April, when OpenAI — an artificial intelligence company that has received billions of dollars of investment from Elon Musk and Microsoft, among others — released DALL-E 2, which can create photorealistic images from text prompts. And while initially only the company’s researchers could use the program, OpenAI has been gradually extending access to people who signed up on a waiting list.

(Google is playing this game too, showing even more vivid images produced by its Imagen system — which, however, is not now open to private use.)

At first, the DALL-E 2 images released were what you might expect with a technology that’s in the hands of technology nerds: Bears programming computers, Darth Vader on the cover of Vogue; a monkey with sunglasses smoking a cigarette and watching the sunset, a painting of tigers wearing VR headsets during the Song dynasty.

But on May 27, Noam Zur, apparently an Israeli who has access to Dall-E2 and identifies himself as “a proud Jewish orthodox,” began tweeting out, under the account name @avocados_ai, “scenes from the Jewish life, demonstrated by avocados.”

He asked the machine to produce such images as “Avocado mom and and avocado daughter, lighting Shabbat candles together” and “Avocado returns to the land of Israel after 2000 years of exile, oil painting” and “A Jewish avocado looking at the night sky, he is looking for the first three stars, an oil painting by Vincent Van Gogh.”

So far, he has tweeted out 48 times, some offering many images in response to the same prompt.

Back to the future of art, and art students: If five years ago you wanted a coffee table or novelty book of avocados living their best Jewish life in a variety of artistic styles, you would have to calculate the cost of commissioning dozens of pieces of art from artists skilled enough to work in a variety of styles. And you would expect to pay each artists for the many hours required to create the work — far fewer hours for artists using digital painting tools like Photoshop than using oils required a generation ago, but many hours nonetheless. It’s hard to imagine enough people paying for a book of Jewish avocado art to make that a profitable venture.

Now, each image is generated in a matter of minutes.

Oh, what brave new world that has such Jewish avocados in it.