The (Robot) Creative Class

Machines have been chipping away at manufacturing jobs for decades, proving themselves faster, cheaper, and more precise than flesh-and-blood line workers. Now engineers are aiming higher than factory drudgery, pivoting from menial to artisanal labor once thought exclusively human. It’s baby steps so far—no need to panic—but according to Moore’s Law, machines’ skills do double every two years. So …

Illustrations by Zohar Lazar

The gags by Scottish computer scientists’ System to Augment Non-speakers’ Dialogue Using Puns (STANDUP) are truly gag-worthy, but better is the University of Washington’s Double Entendre via Noun Transfer (DEviaNT), which knows when to add “that’s what she said” to the end of a sentence with 71.4 percent accuracy. DEviaNT relies on 262 examples of when to use the phrase and 20,700 examples of when to not, which is to say that a lot went into it.

CAI (Creative Artificial Intelligence; pronounced “kay”) is advertising software from a Paris-based company. Pick a product—say, milk. Add an objective, such as “generate awareness,” “create loyalty,” or “increase purchase,” and define the product’s benefits (“healthy”). CAI will then generate a product-specific billboard, like an adorable little girl with a milk mustache and the line “There’s no set age to be healthy.” Days’ worth of human brainstorming and labor knocked out in a few robo-seconds.

Instead of employing gaggles of young lawyers, law firms have achieved recall and precision in document review far beyond the “practical upper bound” of human effort by using programs first designed as spam filters. Robots are also learning to fudge the truth, putting us a step closer to automated trial lawyers: Ron Arkin’s “decepticons” at Georgia Tech offer red herrings, bluffs, and white lies. Nico, a Yale bot, flat-out cheats at rock-paper-scissors.

Emily Howell, a classical-composer bot in California, is so prolific it would take any Mozart or Bach centuries to catch up. Its works have even brought people to tears. And for those who dismiss classical as too math-derived, there’s Shimon, a percussionist bot at Georgia Tech that can improvise jazz and even play with a human ensemble.

Programmers at MIT gave Makr Shakr the ability to produce a googol’s worth (1 x 10100) of drink combinations; by contrast, speakeasy mogul Sasha Petraske’s premier mixologists memorize about 500. But the real impact, its backers say, is its algorithmic objectivity. No more buybacks, sorry, but also no more long lines, favors to prettier faces, snubbed tourists, or eye-rolls when you order that White Russian with vanilla vodka and a splash of Diet Coke.

Developed with Northwestern University’s prestigious Medill School of Journalism, Chicago-based Narrative Science created a computer program that writes basic news articles like sports-game summaries and earnings reports. It already has at least 30 clients, including Forbes and sports heavyweight the Big Ten Network.

For $6,995—$100 extra for pubic hair—New Jersey–based True Companion sells what it bills as “the world’s first sex robot.” Both male (Rocky) and female (Roxxxy) versions are the same size: five foot seven and 120 pounds. Breasts and penises come in customizable cup sizes and lengths, and there are seven personality settings, including Mature Martha (“more of a conversationalist,” says Doug Hines, its creator), Young Yoko (who’s barely 18), S&M Susan, and Frigid Farrah.

In London, the Painting Fool produces “context-free design grammar”—robot imagination, in other words. It simulates mediums to capture a subject’s mood: for example, graphite pencil and muted colors for a sketch called Really Sad. Ibis Hotels in Berlin, London, and Paris have rooms where robots use sensors to scan your sleep and paint your dreams.

Have good intel? Send tips to

The (Robot) Creative Class