Sign In

Communications of the ACM

ACM TechNews

I've Taught Computers to Paint Portraits — and How to Code

View as: Print Mobile App Share:
Simon Colton

"Mathematicians like software to do the boring grunt work," Simon Colton says. "But creative things . . . they like to do themselves"

Credit: Henrik Frydkjaer

Falmouth University professor Simon Colton's area of focus is software that behaves in a manner that would be deemed creative if observed in humans. In an interview, Colton describes his work with software dubbed HR, which is programmed to make its own discoveries.  One such discovery HR made was of the algebraic classification of Latin squares, while another was the independent derivation of Goldblach's conjecture, Colton says.  

Designing software to make discoveries involves feeding it data that you want to learn something about, Colton says. But instead of seeking known unknowns — as with machine learning — it attempts to find unknown unknowns.  "So we teach it how to do general things rather than specifics," he says.  "That contradicts most of what we do in computer science, which is to make sure software does exactly what you want."  

Another program Colton developed, the Painting Fool, creates portraits, and he says mathematicians easily accept computers as creative if they produce great results repeatedly — but persuading artists is another matter altogether.  Colton speculates true computer discovery will only arrive when software becomes capable of self-programming.

From New Scientist
View Full Article


Abstracts Copyright © 2015 Information Inc., Bethesda, Maryland, USA


No entries found

Sign In for Full Access
» Forgot Password? » Create an ACM Web Account