acm-header
Sign In

Communications of the ACM

ACM TechNews

Looking For Art in Artificial Intelligence


Who will make a better dance mix, a computer or a human?

Dartmouth College professors are running a contest to determine whether an algorithm can produce human-quality dance music.

Credit: Annelise Capossela

Dartmouth College professors Michael Casey and Daniel Rockmore this month will explore the potential for algorithms to produce human-quality dance music, sonnets, and short stories as part of their "Turing Tests in the Creative Arts" series.

The dance music contest tasks programmers to build a fun dance set from a predefined library of music, using a single track from the database as the initial random "seed." The software should be able to construct a 15-minute set from this seed, mixing and tweaking choices from the archive, which includes standard annotations of more than 20 features.

Meanwhile, participants in the sonnet and short story competitions must submit self-contained software packages that, upon the input of a common noun phrase, can generate the desired literary output, with the code ideally capable of generating an unlimited number of different works from a single given prompt.

Casey and Rockmore say the test will involve screening the computer-made entries to rule out obvious machine-made creations, while human-generated work will be mixed in with the rest. They will ask a panel of judges to determine whether they think each entry is human- or machine-generated.

Students will score the entries for the dance music contest, with the winner being statistically indistinguishable from the human-generated work.

From The Conversation
View Full Article

 

Abstracts Copyright © 2016 Information Inc., Bethesda, Maryland, USA


 

No entries found