Roboticist trains AI to write fortunes - and things get weird
When Alexander Reben began feeding thousands of inspirational expressions he scraped off the Internet into an algorithm he designed, he wasn't sure what might happen next.
The goal was to train artificial intelligence to create the sort of generic messages that someone might find in fortune cookies.
The result took the artist and MIT-trained roboticist by surprise. Instead of producing the kind of playful and seemingly vaguely perceptive advice known to bring a smile to people's faces, Reben's technology turned dark - and undeniably weird.
About 75 percent of the machine's fortunes, Reben estimates, ended up being "very negative," though often no less hilarious. He calls the results "artificial philosophy."
A few examples:
1. "A friend is a wonderful invention; but, then again."
2. "Being happy is not as serious as it appears."
3. "Success is a powerful excuse."
4. "Politicians broke my little house."
"People think that this sort of computer science is very predictable, mathematical and cut-and-dry, but when you have such a large data set and complicated algorithms, you can get very unexpected outputs," Reben said. "These fortunes also have a beauty and humor that is all of their own, and there's this inherent creativity to a lot of these algorithms, as well."
When a machine-generated fortune strikes Reben's fancy, the Berkeley, Calififornia, resident prints the expression on colorful cotton paper, creating 22-inch prints that are being sold through the Charlie James Gallery in Los Angeles. He's already getting offers for the work, he said.
Using his penchant for playful experimentation, Reben has a knack for highlighting AI's immense potential and revealing that which feels nearly human in nature. Simultaneously, his work lampoons technology, revealing its absurdity and periodic recklessness.
Earlier this year, Reben created a system that allowed a Google Home smart speaker to command a device to fire a pellet gun, generating several hundred thousands views on YouTube.
"OK, Google," he says on the video, "activate gun."
More recently, Reben fed the Quran and Bible into his algorithmic system and now has the ability to generate an infinite number of new religious texts. Wearing a robotic mask, he also recently gave a TEDx talk entirely written by a computer.
Though it might be easy to dismiss Reben's latest project as the nonsensical output of a random algorithm, the artist says that misses the point. What makes the work unique, he argues, isn't the seemingly random forces that form the words but the inevitable human reaction they elicit as we project our own meaning onto the expressions.
It's that interplay, created by the human tendency to see ourselves in our technology, that Reben is interested in exploring. As AI becomes an increasingly common part of our lives, he believes technologists will have reconciled this distinctly human tendency. He predicts that increasingly sophisticated technology may become intolerable if it's not imbued with sufficient human characteristics.
"If we get inside a self-driving car and it does something wrong, for example, we might be angry at it," Reben said. "I actually predict some of the more successful cars in the future will need to have some sort of chauffeur personality to make people feel comfortable, one that can explain why the car did what it did using a voice or a face."
"Neglecting to add a human side to artificial intelligence is a mistake," he added.