TY - GEN
T1 - Investigating Human Priors for Playing Video Games
AU - Dubey, Rachit
AU - Agrawal, Pulkit
AU - Pathak, Deepak
AU - Griffiths, Thomas L.
AU - Efros, Alexei A.
N1 - Funding Information:
We thank Jordan Suchow, Nitin Garg, Michael Chang, Shub- ham Tulsiani, Alison Gopnik, and other members of the BAIR community for helpful discussions and comments. This work has been supported, in part, by Google, ONR MURIN00014-14-1-0671, Berkeley DeepDrive, NVIDIA Graduate Fellowship to DP.
Publisher Copyright:
© 2018 35th International Conference on Machine Learning, ICML 2018. All rights reserved.
PY - 2018
Y1 - 2018
N2 - What makes humans so good at solving seemingly complex video games? Unlike computers, humans bring in a great deal of prior knowledge about the world, enabling efficient decision mak-ing. This paper investigates the role of human priors for solving video games. Given a sample game, we conduct a series of ablation studies to quantify the importance of various priors on human performance. We do this by modifying the video game environment to systematically mask different types of visual information that could be used by humans as priors. We find that removal of some prior knowledge causes a drastic degradation in the speed with which human players solve the game, e.g. from 2 minutes to over 20 minutes. Furthermore, our results indicate that general priors, such as the importance of objects and visual consistency, are critical for efficient game-play. Videos and the game manipulations are available at https : //rach0012. github.io/humanRL-website/.
AB - What makes humans so good at solving seemingly complex video games? Unlike computers, humans bring in a great deal of prior knowledge about the world, enabling efficient decision mak-ing. This paper investigates the role of human priors for solving video games. Given a sample game, we conduct a series of ablation studies to quantify the importance of various priors on human performance. We do this by modifying the video game environment to systematically mask different types of visual information that could be used by humans as priors. We find that removal of some prior knowledge causes a drastic degradation in the speed with which human players solve the game, e.g. from 2 minutes to over 20 minutes. Furthermore, our results indicate that general priors, such as the importance of objects and visual consistency, are critical for efficient game-play. Videos and the game manipulations are available at https : //rach0012. github.io/humanRL-website/.
UR - http://www.scopus.com/inward/record.url?scp=85057277730&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85057277730&partnerID=8YFLogxK
M3 - Conference contribution
AN - SCOPUS:85057277730
T3 - 35th International Conference on Machine Learning, ICML 2018
SP - 2160
EP - 2168
BT - 35th International Conference on Machine Learning, ICML 2018
A2 - Krause, Andreas
A2 - Dy, Jennifer
PB - International Machine Learning Society (IMLS)
T2 - 35th International Conference on Machine Learning, ICML 2018
Y2 - 10 July 2018 through 15 July 2018
ER -