I last spoke to Keret, my favorite Israeli writer, a dozen days after the Hamas terror attack, when he was experiencing his first writer’s block in 37 years and rushing around the country doing impromptu readings for the traumatized. I reached back out as the war approached its five-month mark because he always provides me with poetic insight to Israel’s psyche.
We spoke the morning of the deadly aid-convoy disaster in Gaza, as he was trying to finalize the cover for his seventh short story collection, Autocorrect, which publishes in Hebrew at the end of May (the English versions usually come a year later). He was preparing for a big event honoring his mother, who died four years ago, and is still doing readings for soldiers and evacuees.
He is also back teaching at Ben Gurion University — and working with researchers at Tel Aviv University on a project, as he put it, “trying to teach an AI how to write better fiction.”
What?
“We try to work on concentrated issues, like let’s say metaphors,” Keret attempted to explain. “Now I’m working with him on openings. I give him a premise for a story, and then I ask him to write the first paragraph. Then I read it, and I say to him, ‘Yeah, you know, but you’re telling the entire story in the paragraph.’ Or, ‘This is not interesting.’ Or this is unrelated, or this is cliche. I try sometimes to formulate rules that will help him, just to kind of make him do a little bit less bad.”
How’s that going?
“It’s half frustrating,” Keret said. “It’s not something that I’m doing and say, ‘Oh it’s amazing, it’s going to change the world. Half of the time I’m saying, ‘Wow, this was so vain of me.’”
I wondered why he’d want to help AI get better at writing fiction in the first place. Isn’t that exactly the kind of thing we think only humans can do well? And wouldn’t it, you know, be bad for business?
Somehow this led us to space travel.
“Once we said, ‘Oh if we can improve space travel, we can get further, we can know more.’ And then we say, ‘Ah, we have space travel so we can sell it to billionaires and they fly to space and do a selfie,’” Keret said. “There is something about our interaction with AI that is so self-indulgent. This thing can solve the biggest problem in the world, but the entire way that we conceptualize it is if it can draw a cute picture of a squirrel, or of our neighbor on fire.
“The more I deal with AI the more it gets me thinking of humanity,” he continued. “I thought to myself: Maybe the trick will be to help the AI understand the humanity better, and then it will make the hole bigger so we will be able to crawl through. It will leave us more space to be ourselves or not to change.”
Right, so how’s that going?
“My students are much, much more talented,” Keret said.
“Writing is an attempt to bring an individual thought to people that don’t think the same way,” he explained. “There is something about the way that the AI works, it’s built on statistics, it’s kind of wisdom of the crowd. So all the time it kind of says the thing that the biggest number of people would say. The effect of that is really kind of uninspired.”
Suddenly, it seemed we were back to talking about the war.
“Usually, the way you identify somebody who is missing is by her scars or by his birthmarks,” he noted. “Those things, they are the thing that the AI doesn’t have, because they are a mistake.
“The idea was that technology would help us have more free time and be ourselves. What happened is it pulls us away from being ourselves,” he explained. “AI can take something positive and turn it into something addictive. If I like to see sunsets, I can only see one a day. But with technology, I can see sunsets on loops. All the balancing mechanisms are being thrown away.”
