Editorial: 鈥�To make your star quality real, you need a roadmap鈥�
WHETHER you鈥檙e a software developer, an engineer, or just designing a family greeting card, chances are you鈥檝e been stuck on a creative project at one point or another. Wouldn鈥檛 it have been nice to call on the creative juices of a few dozen strangers 鈥� or even artificial intelligence 鈥� to help you overcome your block?
You may soon be able to. Researchers are building new tools to harness the crowd for creative pursuits 鈥� and even building in algorithms that can take human ideas and 鈥渆volve鈥� them in directions that people may never have thought of.
Advertisement
Anand Kulkarni, co-founder of , a crowdsourcing website based in Berkeley, California, is preparing to launch Wish, an interface that lets users of desktop software access crowd intelligence from within an application. So if you want to improve an image or document you are working on, people in the crowd will tell you what they think. 鈥淏ecause it is embedded in the interface of a creative tool like Photoshop, a word processor, or something more, Wish provides a natural amplification of the user鈥檚 creative ability,鈥� says Kulkarni.
聯If you want to improve an image or document, people in the crowd will tell you what they think聰
Wish will initially plug into three web apps: , which is already up and running, lets web-based editors suggest improvements to your writing. When ready, 鈥淗ack鈥� will let developers improve your software code and 鈥淪ketch鈥� will allow you to submit a rough digital drawing and get back a finished work.
Since most users will not want a bunch of novices mucking around with their project, Wish does something most crowdsourcing tools don鈥檛 鈥� it asks the crowd itself to find skilled experts, and vets them based on past work and credentials. It then routes tasks directly to the experts, who can select which projects they would like to work on for payment.
Meanwhile, at Carnegie Mellon University in Pittsburgh, Pennsylvania, and Jeffrey Nickerson of Stevens Institute of Technology in Hoboken, New Jersey, are using a blend of crowds and artificially intelligent software to enhance creativity.
They set members of a crowd the task of inventing a novel device. Once designs have been submitted, an evolutionary algorithm develops the ideas. Mimicking natural selection, the algorithm randomly mutates aspects of the designs, trying to hit upon devices that better meet pre-set parameters and goals. Evolutionary algorithms are often used to help improve things like aircraft wings or radio antennas in ways that humans might not have conceived of.
Yu and Nickerson decided to build on this ability by throwing designs that have gone through algorithms to a new human crowd to assess the usefulness of an invention. In a test of their idea, they farmed out a task to design a small chair for children to a group of 60 paid workers on Amazon鈥檚 Mechanical Turk. A second Turker crowd fed the ideas they thought best into an algorithm that mixed and improved them further. This cycle repeated several times with new Turker groups, and then a final crowd chose the most novel yet practical chair idea. 鈥淥ne chair was designed like a crescent and a smiley face; another was designed in the shape of a bear,鈥� says Yu.
鈥淲hat they have done is pull out the new and creative bits and then pushed the crowd in more interesting and novel directions,鈥� says at the University of Copenhagen in Denmark.
There are limitations to where crowd creativity is safe to go, Yu admits. 鈥淵ou wouldn鈥檛 want to build a submarine through a crowdsourcing project 鈥� although the result would be interesting.鈥�
Eye-catching design
Evolutionary algorithms need directing so as not to deviate from a designer鈥檚 goals (see main story). But the process can be slow when commands need to be typed, so Hod Lipson and colleagues at Cornell University in Ithaca, New York, .
This allows designers to gaze at a screen full of different designs as an algorithm comes up with them. The software throws out designs that do not catch the user鈥檚 eye, and instead makes further iterations of the ones being looked at. 鈥淯sers can produce novel, interesting shapes without a keyboard,鈥� the team reports.
The idea is to build a system that evolves 3D-printable objects 鈥� with little more than a glance of approval.
This article appeared in print under the headline 鈥淐reativity on tap鈥�
