Skip to main content

We asked a neural network to bake us a cake. The results were...interesting.

[ad_1]


When computers try to imitate humans, they often get confused. But simulated brain cells in so-called neural networks can mimic our problem-solving skills. An AI will look at a dataset, figure out its governing rules, and use those instructions to make something new. We already employ these bots to recognize faces, drive cars, and caption images for the blind. But can a computer cook? I fed a neural network thousands of recipes and asked it to whip up something of its own. It spat out an...interesting dish.



Amnesiac AI



To keep processing fast, the network recalls only 65 characters at a time. It adds cocoa just before it would otherwise forget we’re making “chocolate.” Hopefully it forgets that black puddings usually feature blood.



Word salad



Title, category, ingredients, directions. It gets this format every time because of consistency across all the recipes we pulled from. But my AI hasn’t seen as many examples of how to use rare ingredients like sesame.



Date holy



Sounds like frosting! We’re staying in dessert territory. But what’s with the name? Capital letters are tough; they’re treated as unrelated to lowercase. A neural net has to learn these from scratch with few examples.


Sweet defeat



The ingredient list has already passed out of memory. Chocolate is just a lucky guess pulled from your average cookbook. “Until golden brown” could mean sweet or savory, and ambiguity confuses the network. Garlic? Game over.



Feedback



The confused network is spitting out random words—creating more confusion, leading to more random words. But it remembered to close its parenthesis. There’s probably a neuron solely devoted to parentheses.



Yum?



Words like “frost” or “serve” can cue a ­network to finish; many recipes in our dataset end this way. Ours just happened to give up—it could have kept going forever. We were making cake, right? Let’s say it’s cake.





This article was originally published in the Spring 2018 Intelligence issue of Popular Science.




[ad_2]

Written By Janelle Shane

Comments

Popular posts from this blog

Ice technicians are the secret stars of the Winter Olympics

[ad_1] The emphasis of this year's two-week-long Winter Olympic Games has been placed squarely on the Olympians themselves. After all, the stated purpose of the international competition is to bring together the world’s greatest athletes in a nail-biting competition across fifteen different winter sports. But before the curlers, skiers, and skaters even arrived in Pyeongchang, South Korea, the Olympians of the ice technician world were already a few weeks deep in a competition of their own. Mark Callan of the World Curling Federation and Markus Aschauer of the International Bobsleigh and Skeleton Federation both say they’re hoping to make the best ice the Winter Olympics have ever seen. To transform the barren concrete jungle of existing tracks and arenas into an ice- and snow-covered wonderland is an enormous undertaking. And it takes a keen understanding of the physics and chemistry that keeps frozen precipitation pristine. Curling Callan has been making and maintaining ic...

With Operation Popeye, the U.S. government made weather an instrument of war

[ad_1] It was a seasonably chilly afternoon in 1974 when Senators Claiborne Pell, a Democrat from Rhode Island, and Clifford Case, a Republican from New Jersey, strode into the chambers of the Senate Committee on Foreign Relations for a classified briefing. While the meeting was labeled “top secret,” the topic at hand was rather mundane: They were there to discuss the weather. More specifically, Pell, the chairman of the now-defunct subcommittee for Oceans and International Environment, and his colleague were about to learn the true extent of a secret five-year-old cloud seeding operation meant to lengthen the monsoon season in Vietnam, destabilize the enemy, and allow the United States to win the war. Though it cycled through several names in its history, "Operation Popeye" stuck. Its stated objectives—to ensure Americans won the Vietnam War—were never realized, the revelation that the U.S. government played God with weather-altering warfare changed history. The...

University supercomputers are science's unsung heroes, and Texas will get the fastest yet

[ad_1] Supercomputers are powerful machines with great names—Blue Waters, Bridges, Jetstream, Comet. But a new one will soon be joining that list: Frontera. The $60 million machine will live at the University of Texas at Austin and is scheduled to come online next year. “It will be the fastest machine ever deployed at a university in the US,” says Dan Stanzione, the executive director of the Texas Advanced Computing Center. With supercomputers, the title of fastest is a moving target—what’s perhaps more important is not the exact ranking, but that they’re available for researchers to use in the first place. Right now, the fastest supercomputer in the world is called Summit, and it’s at the Oak Ridge National Laboratory in Tennessee, part of the Department of Energy (DOE), and is specifically tailored for AI. But supercomputers located not at government labs but at universities—like Frontera and its ilk—play a crucial role in the ever...