Christoph Salge and Alan Blair and Alex J. Champandard and Michael Cook and Raluca D. Gaina and Diego Perez-Liebana and Jacob Shrum and Emily Short and Vanessa Volz
This abstract reports both on a talk and on a subsequent working group with the same title. The initial presentation was focused on the GDMC AI Settlement Generation Challenge in Minecraft. This is a competition where participants write an algorithm that can generate an “interesting” settlement for a given, previously unknown Minecraft map. The general idea behind the competition is to put some focus of creativity and co-creativity as part of intelligence, and consequently as part of artificial intelligence. Currently, the main focus is on creating a settlement that fits into, and reacts to the environment given by an input map. The output should be similar to what a human could do. Further down the line there are plans to further extend the competition, to have AIs complete partially built settlements, and possibly do so from an embodied perspective. Currently, the AI interact with the map as a 3D representation. While the GDMC challenge hopes to stimulate research in this direction, it also aims to a.) solicit some ideas and approaches from the general public and b.) offer a joint framework where different technical approaches, such as wave form collapse or PCGML can be compared with each other.
Cite this work
@inproceedings{salge19511, author= {Christoph Salge and Alan Blair and Alex J. Champandard and Michael Cook and Raluca D. Gaina and Diego Perez-Liebana and Jacob Shrum and Emily Short and Vanessa Volz}, title= {{Procedural Content Generation for Minecraft}}, year= {2018}, journal= {{Artificial and Computational Intelligence in Games: Revolutions in Computational Game AI (Dagstuhl Seminar 19511)}}, volume= {9}, number= {12}, pages= {67--114}, editor= {Jialin Liu and Tom Schaul and Pieter Spronck and Julian Togelius}, abstract= {This abstract reports both on a talk and on a subsequent working group with the same title. The initial presentation was focused on the GDMC AI Settlement Generation Challenge in Minecraft. This is a competition where participants write an algorithm that can generate an “interesting” settlement for a given, previously unknown Minecraft map. The general idea behind the competition is to put some focus of creativity and co-creativity as part of intelligence, and consequently as part of artificial intelligence. Currently, the main focus is on creating a settlement that fits into, and reacts to the environment given by an input map. The output should be similar to what a human could do. Further down the line there are plans to further extend the competition, to have AIs complete partially built settlements, and possibly do so from an embodied perspective. Currently, the AI interact with the map as a 3D representation. While the GDMC challenge hopes to stimulate research in this direction, it also aims to a.) solicit some ideas and approaches from the general public and b.) offer a joint framework where different technical approaches, such as wave form collapse or PCGML can be compared with each other.},
}