The DigiLit prize competition encourages the creation of algorithms able to produce "human-level" short story writing that is indistinguishable from an “average” human effort. A 2016 competition revealed just how difficult this is. For 2017, the organizers will give a prize to contestants able to create algorithms that best complete a short story. That is, given a roughly 1000-2000 word story prompt that has no ending, contestant must generate a 300-500-word conclusion to the story.
- Entries must be submitted by 8:00AM EST, Jun 1, 2017. Contestants will submit their work to a virtual machine running a Linux operating system. Participants will be responsible for getting their software to run on the platform (although the organizers will do their best to help if problems arise). See the description of the testing environment for details.
- The virtual machine should be able to run the code and generate a story ending of at most 500 words within 24 hours using less than 16GB of RAM and less than 50GB of disk space.
- The initial story text will be provided in plaintext form.
- We prefer that submissions include source code under an Open Source License. The program must be completely self-contained and require no connectivity to the Internet.
- Entries will be tested and a set of finalists will be determined.
- Finalists’ conclusions will be tested against human-generated conclusions to the same short story. That is, the “conclusion” output of the program will be attached to the body of its short story, and the full hybrid story will be added to a collection of stories all with the same body and with different human-written conclusions. Judges will be asked to identify the human-generated endings and the machine-generated endings.
- Among the finalists, any programs that “pass” the Turing Test—i.e., those that create conclusions that the judges believe to be human-generated—will receive a $2000 cash prize.
- Judges will give prizes for outstanding creative efforts at their discretion.